JP3551666B2 - Moving picture coding method and moving picture coding apparatus - Google Patents

Description

である。
【０１５２】
また、動画像符号化回路１２３には、例えば上述の図９に示した構成のものが用いられる。また、符号化難易度測定回路１２６には、例えば上述の図１０に示した構成のものが用いられる。
【０１５３】
次に、本発明の第４の構成例の動画像符号化装置を図２１に示す。
【０１５４】
この図２１に示す第４の構成例において、上述した第３の構成例の動画像符号化装置との大きな違いは、プリフィルタ制御回路１９７に入力する符号化難易度ｄ（ｄ＿ｃｕｒｒｅｎｔ ）として、動きベクトル検出回路１９６から出力される予測残差、すなわち前述の式（１） で計算される、現在のブロックの信号Ａｉｊと任意の動きベクトルにより参照されるブロックの信号Ｆｉｊの差の絶対値の和Ｅｆ を用いている点である。
【０１５５】
この図２１において、入力端子１９０から入力されるディジタル動画像信号Ｓ１００は、フレームメモリ群１９１に送られ、記憶される。このフレームメモリ群１９１に記憶された画像データＳ１０３を用いて、動きベクトル検出回路１９６は、フレーム間の動きベクトルを前記第３の構成例で説明したにように検出する。
【０１５６】
動きベクトル検出回路１９６では、入力画像の小ブロック毎に動きベクトル信号ＭＶと上述の式（１） から計算される予測残差Ｅｆ を計算し、動きベクトル信号ＭＶを動画像符号化回路１９３とプリフィルタ制御情報記憶回路１９５に入力する。
【０１５７】
画像処理タイプ判定回路１９４では、入力された画像データの処理番号ｐｉｃｔ＿ｎｕｍｂｅｒより、現在処理中の画像がフィルタ特性画像であるか非フィルタ特性画像であるかの判定を行う。現在処理中の画像がフィルタ特性画像、例えばＰピクチャ、であったならば、回路切替え信号ｐｉｃｔ＿ｔｙｐｅ（１）をフィルタ制御情報記憶回路１９５と動きベクトル検出回路１９６に入力する。
【０１５８】
回路切替え信号ｐｉｃｔ＿ｔｙｐｅ（１） を入力された動きベクトル検出回路９６では、予測残差Ｅｆ を、その小ブロックの符号化難易度ｄ（ｄ＿ｒａｗ）として出力する。
【０１５９】
符号化難易度平滑化回路１０９では、近接する小ブロック間で、フィルタ係数が極端に異なることを避けるために、入力したｄ（ｄ＿ｒａｗ ）をフレーム内で平滑化し、平滑化符号化難易度情報ｄ（ｄ＿ｃｕｒｒｅｎｔ ）を計算する。符号化難易度平滑化回路１９９での平滑化処理は第３の構成例と同様に行う。プリフィルタ制御回路１９７へは、現在の符号化対象の小ブロックの符号化難易度ｄ（ｄ＿ｃｕｒｒｅｎｔ） が入力される。また、プリフィルタ制御情報記憶回路１９５には、過去に入力された小ブロックで使われたプリフィルタ係数が記憶されている。ここで、このプリフィルタ制御情報記憶回路１９５には、現在の小ブロックのフレーム上での位置を表すブロックアドレスｍｂ＿ａｄｄｒｅｓｓ と現在の小ブロックの動きベクトル信号ＭＶ及び、回路切替え信号ｐｉｃｔ＿ｔｙｐｅ（１）が入力される。
【０１６０】
このプリフィルタ制御情報記憶回路１９５では、これらに基づいて現在の小ブロックがフレーム間予測で参照する小ブロックで使用されたローパスプリフィルタの通過帯域制限を指定するパラメータとしてフィルタ係数ｋ＿ｒｅｆ（１）を読み出し、これをプリフィルタ制御回路１９７へ入力する。
【０１６１】
なお、フィルタ係数ｋ＿ｒｅｆ（１）を読み出す場合は、ブロックアドレスｍｂ＿ａｄｄｒｅｓｓ だけを使用して、フレーム間予測で参照するフレーム上の同じブロックアドレスｍｂ＿ａｄｄｒｅｓｓ の位置にある小ブロックで使用されたローパスプリフィルタの通過帯域制限を指定するパラメータをフィルタ係数ｋ＿ｒｅｆ として読み出しても良い。この場合は、プリフィルタ制御情報記憶回路１９５へ動きベクトル信号ＭＶを入力する必要はない。
【０１６２】
プリフィルタ制御回路１９７は、上記符号化難易度ｄとフィルタ係数ｋ＿ｒｅｆ（１）が入力されると、これらに基づいて現在の小ブロックに使用するローパスプリフィルタの通過帯域制限を指定するパラメータとしてフィルタ係数ｋ＿ｃｕｒｒｅｎｔ を生成出力し、これをプリフィルタ１９２へ入力する。ここで、当該プリフィルタ１９７の特性は、例えば前述した図２４のような特性とする。したがって、ここでは、上記フィルタ係数ｋ＿ｃｕｒｒｅｎｔ が小さいほど、通過帯域の狭いローパスフィルタ特性となる。
【０１６３】
なお、プリフィルタ１９２の特性は、図２４に示した特性のローパスフィルタに限らず、これ以外の特性をもつローパスフィルタであってもよい。
【０１６４】
プリフィルタ制御回路１９７で計算されたフィルタ係数ｋ＿ｃｕｒｒｅｎｔ は、また、プリフィルタ制御情報記憶回路１９５へ入力されて記憶され、これが未来に入力される小ブロックに使うプリフィルタ係数を決定する時の参照値ｋ＿ｒｅｆ として利用されることになる。
【０１６５】
また、画像処理タイプ判定回路１９４において、入力されたｐｉｃｔ＿ｎｕｍｂｅｒ より、現在処理中の画像が非フィルタ特性画像、例えばＩ、Ｂピクチャ、であると判定した場合、回路切替え信号 ｐｉｃｔ＿ｔｙｐｅ（２） をフィルタ制御情報記憶回路１９５と動きベクトル検出回路１９６に入力し、現在の小ブロックのフレーム上での位置を表すブロックアドレスｍｂ＿ａｄｄｒｅｓｓ をプリフィルタ制御情報記憶回路１９５と内挿補間回路２００に入力する。
【０１６６】
回路切替え信号 ｐｉｃｔ＿ｔｙｐｅ（２） を入力された動きベクトル検出回路１９６では、予測残差Ｅｆ の出力を行わない。
【０１６７】
現在の小ブロックのブロックアドレスｍｂ＿ａｄｄｒｅｓｓ と現在の小ブロックの動きベクトルＭＶ及び、回路切替え信号 ｐｉｃｔ＿ｔｙｐｅ（２） を入力されたプリフィルタ制御情報記憶回路１９５は現在の小ブロックがフレーム間予測で参照する未来参照画像と過去参照画像の小ブロックで使用されたローパスプリフィルタの通過帯域制限を指定するパラメータとしてのフィルタ係数ｋ＿ｒｅｆ（２）、ｋ＿ｒｅｆ（３）と時間的に近接する画像の空間的に同じ位置の小ブロックで使用されたローパスプリフィルタの通過帯域制限を指定するパラメータとしてのフィルタ係数ｋ＿ｒｅｆ（４）を読み出し、内挿補間回路２００に入力する。
【０１６８】
なお、上記フィルタ係数ｋ＿ｒｅｆ（２）、ｋ＿ｒｅｆ（３）を読み出す場合には、ブロックアドレスｍｂ＿ａｄｄｒｅｓｓ だけを使用して、フレーム間予測で参照するフレーム上の同じブロックアドレスｍｂ＿ａｄｄｒｅｓｓ の位置にある小ブロックで使用されたローパスプリフィルタの通過帯域制限を指定するパラメータを上記フィルタ係数ｋ＿ｒｅｆ（２），ｋ＿ｒｅｆ（３）として読み出すようにしても良い。この場合は、プリフィルタ制御情報記憶回路１９５へ動きベクトル信号ＭＶを入力する必要はない。
【０１６９】
現在の小ブロックの動きベクトル信号ＭＶ及び、パラメータとしてのフィルタ係数ｋ＿ｒｅｆ（２），ｋ＿ｒｅｆ（３），ｋ＿ｒｅｆ（４）を入力された内挿補間回路２００は、過去参照画像のフィルタ係数ｋ＿ｒｅｆ（２）と未来参照画像のフィルタ係数ｋ＿ｒｅｆ（３）を基に内挿補間を行い、更に時間的に近接する画像の空間的に同じ位置の小ブロックのフィルタ係数ｋ＿ｒｅｆ（４）を用いて現在の小ブロックのフィルタ係数ｋ＿ｃｕｒｒｅｎｔ を生成出力し、これをプリフィルタ２９２へ入力する。
【０１７０】
ここで、現在の小ブロックと時間的に近接する画像の空間的に同じ位置の小ブロックで使用されたフィルタ係数ｋ＿ｒｅｆ（４）を用いず、現在の小ブロックがフレーム間予測で参照する小ブロックで使用されたフィルタ係数ｋ＿ｒｅｆ（２）、ｋ＿ｒｅｆ（３）のみを読み出して現在の小ブロックのフィルタ係数ｋ＿ｃｕｒｒｅｎｔ を生成出力すようにしても良い。
【０１７１】
また、非フィルタ特性画像が例えばＭＰＥＧ方式のＩピクチャーのように、フレーム間予測を行わない画像であった場合には、時間的に前後するフィルタ特性画像の空間的に同じ位置の小ブロックのフィルタ係数を過去参照画像のフィルタ係数ｋ＿ｒｅｆ（２）と未来参照画像のフィルタ係数ｋ＿ｒｅｆ（３）として読み出し、内挿補間を行う。この場合は時間的に近接する画像の空間的に同じ位置の小ブロックのフィルタ係数ｋ＿ｒｅｆ（４）は用いない。
【０１７２】
また、Ｉピクチャーにおいても、動き予測を行っている場合には、Ｉピクチャーもフィルタ特性画像として用いても良い。ここで、当該プリフィルタ１９２の特性は、例えば前述した図２４のような特性とする。したがって、ここでは、フィルタ係数ｋ＿ｃｕｒｒｅｎｔ が小さいほど、通過帯域の狭いローパスフィルタ特性となる。
【０１７３】
内挿補間回路で計算されたフィルタ係数ｋ＿ｃｕｒｒｅｎｔ は、また、プリフィルタ制御情報記憶回路１９５へ入力されて記憶され、これが未来に入力される小ブロックに使うプリフィルタ係数を決定する時の参照値（フィルタ係数ｋ＿ｒｅｆ ）として利用されることになる。
【０１７４】
以上のようにして、適応的にプリフィルタ１９２のフィルタ係数を決定し、符号化画質の主観的印象が良く、かつ符号化効率の良い符号化方式を実現する。
【０１７５】
フィルタ係数ｋ＿ｃｕｒｒｅｎｔ の計算方法は、第１の構成例において図５を用いて説明した計算例と同様である。
【０１７６】
プリフィルタ１９２は、現在の小ブロックに対して、プリフィルタ係数ｋ＿ｃｕｒｒｅｎｔ で指定されるローパスフィルタ処理を施して、処理画像信号Ｓ１０２を出力する。
【０１７７】
当該処理画像信号Ｓ１０２とその動きベクトル信号ＭＶは、動画像符号化回路１９３へ入力され、所定のフレーム間予測符号化処理（例えばいわゆるＭＰＥＧ方式の符号化処理）が施され、この符号化ビットストリームＳ１０４が出力端子１９８から出力される。動画像符号化回路１９３の構成は、第２の構成例において図９を用いて説明したものと同様である。
【０１７８】
このような第３の構成例の動画像符号化装置では、フレーム間予測符号化を施す動画像信号に対するプリフィルタ処理の際のフィルタ特性を、画面を分割する小ブロック単位の符号化難易度ｄに応じて適応的に変更可能としているので、視覚特性を考慮して、画面内の局所的に速い動きや複雑な絵柄であり符号化難易度の大きい部分では通過帯域の比較的狭いローパスフィルタを選択でき、一方、遅い動きや平坦な絵柄であり符号化難易度の小さい部分では通過帯域の比較的広いローパスフィルタを選択できる。
【０１７９】
また、ＭＰＥＧ方式のように画像毎に符号化難易度の異なる符号化方式を考慮し、連続する画像間で極端にプリフィルタ係数が異なることがないようにするために、符号化難易度を参照する画像を特定している。
【０１８０】
更に、第４の構成例の動画像符号化装置では、上記小ブロックの符号化難易度ｄだけでなく、現在の符号化対象の小ブロックがフレーム間予測で参照する小ブロック及び、近接する画像の空間的に同じ位置の小ブロックで使われたプリフィルタ制御情報をも合わせて考慮しており、フレーム間予測符号化効率が低下しないように、上述のように選択されたフィルタ特性を修正して、最終的に現在の符号化対象の小ブロックに対するプリフィルタ特性を決定するようにしているので、ローパスプリフィルタ処理された動画像は、従来よりも、符号化画質の主観的印象が良く、かつ符号化効率の良いものとすることができる。
【０１８１】
ここで、上述した第１乃至第４の構成例の動画像符号化装置での符号化により得られた符号化ビットストリームは、信号記録媒体に記録されたり、伝送路を介して伝送されることになる。
【０１８２】
図２２には、信号記録媒体の一例として光ディスク７０４を用いた例について説明する。この図２２において、端子７００には、上記符号化ビットストリームと、量子化スケール等の後の復号化に必要な情報とからなるデータ列が供給される。このデータ列は、ＥＣＣエンコーダ７０１によってエラーコレクションコードが付加され、変調回路７０２に送られる。当該変調回路７０２では上記ＥＣＣエンコーダ７０１の出力に対して、所定の変調処理、例えば８−１４変調等の処理を施す。この変調回路７０２の出力は記録ヘッド７０３に送られ、当該記録ヘッド７０３にて光ディスク７０４に記録される。
【０１８３】
なお、図２２の例では、信号記録媒体として光ディスクを例に挙げたが、磁気テープ等のテープ状記録媒体や、ハードディスクやフレキシブルディスク等の磁気ディスク媒体、ＩＣカードや各種メモリ素子等の半導体記憶媒体等の信号記録媒体に対して、本発明装置にて符号化した信号を記録することも可能である。また、光ディスクとしては、ピットによる記録がなされるディスクや、光磁気ディスクの他に、相変化型光ディスクや有機色素型光ディスク、紫外線レーザ光により記録がなされる光ディスク、多層記録膜を有する光ディスク等の各種のディスクを用いることができる。
【０１８４】
【発明の効果】
以上の説明で明らかなように、本発明においては、入力動画像を少なくとも１画素からなる小ブロックに分割し、所定の方法により小ブロックの符号化難易度を計算し、この小ブロックの符号化難易度と当該小ブロックが画像間予測で参照する小ブロックで使われたフィルタ制御情報とに基づいて、当該小ブロックに対して適応的にローパスフィルタ処理を施す際のフィルタ特性を決定することにより、ローパスプリフィルタ処理された動画像は、従来よりも、符号化画質の主観的印象が良く、かつ符号化効率の良いものとすることが可能となる。すなわち、本発明によれば、フレーム間予測符号化回路の前段にプリフィルタを置く構成の動画像符号化装置において、フレームの中で部分的に、ローパスプリフィルタの特性を変更できるとともに、予測画像間でお互いの画像信号の帯域が異なることによる符号化効率の低下を小さくすることが可能である。
【０１８５】
また、本発明においては、入力動画像を少なくとも１画素からなる小ブロックに分割し、入力画像をフィルタ特性画像と非フィルタ特性画像に分類し、所定の方法によりフィルタ特性画像の小ブロックの符号化難易度を計算し、フィルタ特性画像においては、この小ブロックの符号化難易度と当該小ブロックが画像間予測で参照する小ブロックで使われたフィルタ制御情報とに基づいて、当該小ブロックに対して適応的にローパスフィルタ処理を施す際のフィルタ特性を決定し、非フィルタ特性画像においては、過去の参照画像と未来の参照画像とから所定の方法により内挿補間を用いて、当該小ブロックに対して適応的にローパスフィルタ処理を施す際のフィルタ特性を算出することにより、ローパスプリフィルタ処理された動画像は、従来よりも、符号化画質の主観的印象が良く、かつ符号化効率の良いものとすることが可能となる。すなわち、本発明によれば、フレーム間予測符号化回路の前段にプリフィルタを置く構成の動画像符号化装置において、フレームの中で部分的に、ローパスプリフィルタの特性を変更できるとともに、予測画像間でお互いの画像信号の帯域が異なることによる符号化効率の低下を小さくすることが可能である。
【図面の簡単な説明】
【図１】本発明の動画像符号化装置の第１の構成例を示すブロック回路図である。
【図２】フレームと小ブロックの関係を示す図である。
【図３】プリフィルタ制御方法の一例を示すフローチャートである。
【図４】符号化難易度の平均値ｄ＿ａｖｅとプリフィルタ係数の代表値ｋ＿ｇｏｐの対応関係を示す図である。
【図５】現在の小ブロックのフレーム間予測で参照される小ブロックで使われたプリフィルタ特性を考慮して、プリフィルタ係数を計算する一例を示す図である。
【図６】符号化難易度ｄ＿ｃｕｒｒｅｎｔとローパスプリフィルタ係数ｋ＿ｃｕｒｒｅｎｔとの対応関係を示す図である。
【図７】プリフィルタ係数ｋ＿ｃｕｒｒｅｎｔの計算方法の一例を示すフローチャートである。
【図８】ローパスプリフィルタ係数ｋ＿ｃｕｒｒｅｎｔの計算例を示す図である。
【図９】本発明構成例の動画像符号化装置内の動画像符号化回路の構成例を示すブロック回路図である。
【図１０】本発明構成例の動画像符号化装置内の符号化難易度測定回路の構成例を示すブロック回路図である。
【図１１】本発明の動画像符号化装置の第２の構成例を示すブロック回路図である。
【図１２】ＭＰＥＧ方式の予測構造及び各ピクチャタイプの構成を示す図である。
【図１３】本発明の動画像符号化装置の第３の構成例を示すブロック回路図である。
【図１４】ＭＰＥＧ方式における表示順序と符号化順序を示す図である。
【図１５】フィルタ特性画像と非フィルタ特性画像のフレームと小ブロックの関係を示す図である。
【図１６】プリフィルタ制御方法の一例を示すフローチャートである。
【図１７】非フィルタ特性画像において、現在の小ブロックのフレーム間予測で参照される小ブロックで使われたプリフィルタ特性を用いて内挿補間により、プリフィルタ係数を計算する一例を示す図である。
【図１８】プリフィルタ係数ｋ＿ｃｕｒｒｅｎｔ の計算方法の一例を示すフローチャートである。
【図１９】非フィルタ特性画像のローパスプリフィルタ係数ｋ＿ｃｕｒｒｅｎｔ の計算例を示す図である。
【図２０】プリフィルタの構成例を示す図である。
【図２１】本発明の動画像符号化装置の第４の構成例を示すブロック回路図である。
【図２２】信号記録媒体の一例として光ディスクに符号化ビットストリームを記録するための構成を示すブロック回路図である。
【図２３】従来の動画像符号化装置の構成例を示すブロック回路図である。
【図２４】プリローパスフィルタの係数を示す図である。
【図２５】プリフィルタ係数と符号化難易度の関係を示す図である。
【符号の説明】
２１，９１，１２１，１９１ フレームメモリ群、２２，９２，１２２，１９２ プリフィルタ、２３，９３，１２３，１９３ 動画像符号化回路、２４，９６，１２４，１９６ 動きベクトル検出回路、２５，９５，１２５，１９５ プリフィルタ制御情報記憶回路、２６，１２６ 符号化難易度測定回路、２７，９７，１２７，１９７ プリフィルタ制御回路、３０１ ローパスフィルタ、３０２，３０３，３０４ 演算器[0001]
TECHNICAL FIELD OF THE INVENTION
The present invention records a moving image signal on a recording medium such as a magneto-optical disk or a magnetic tape, or transmits a moving image signal from a transmission side via a transmission path, such as a video conference system, a video phone system, or a broadcasting device. The present invention relates to a moving image encoding method and apparatus suitable for use in transmitting to a receiving side.
[0002]
[Prior art]
For example, in a system for transmitting a moving image signal to a remote place, such as a video conference system or a video telephone system, the moving image signal is digitized in order to realize high quality transmission. I have. Also, when a moving image signal is recorded on a signal recording medium, a digitized moving image signal is similarly recorded.
[0003]
Here, since a digitized moving image signal has a huge amount of data, data is often encoded (compressed) when recording or transmitting. However, when a moving image signal is encoded (compressed), some deterioration in image quality is inevitable. The image quality degradation includes, for example, block-like distortion and ringing (mosquito noise) in a contour portion. Conventionally, in order to reduce such image quality deterioration, a filtering process is often performed on an input image signal before encoding using a pre-filter (for example, a low-pass filter).
[0004]
FIG. 23 shows an example of a conventional moving picture signal encoding apparatus using the above prefilter.
[0005]
In the encoding device shown in FIG. 23, a digital moving image signal S210 input from a terminal 201 is subjected to a filtering process in a pre-filter circuit 202. The processed image signal S211 that has been subjected to the filtering processing is input to the encoding circuit 205 in the next stage.
[0006]
The encoding circuit 205 encodes a digital video signal using a hybrid encoding method combining motion compensation inter-frame prediction and transform encoding such as DCT (discrete cosine transform). That is, the encoding circuit 205 performs inter-frame / intra-adaptive prediction on the input signal S211 and extracts the prediction error signal S212 by the calculator 206. Further, the DCT unit 207 performs two-dimensional DCT on the prediction error signal. Then, the quantizer 208 quantizes the calculated DCT coefficient, and outputs an encoded output signal S215 from the terminal 220. Here, the quantizer 208 determines a quantization step size such that the bit rate of the output signal S215 is constant.
[0007]
On the other hand, the output signal S 215 is inversely quantized by the inverse quantizer 209, inverse DCT-processed by the inverse DCT unit 210 to restore the prediction error signal, and further predicted by the arithmetic unit 211 to the prediction error signal. The image signals are added and local decoding is performed. The decoded image signal S213 obtained by this addition is stored in the frame memory 212.
[0008]
In general, in the encoding circuit 205, the prediction residual signal S212 increases as the pattern of the input moving image becomes more complicated or the motion thereof increases. At this time, if coarse quantization is used in the quantizer 208 in order to suppress the amount of information generated by encoding to a predetermined bit rate, visually noticeable block distortion occurs and the subjective image quality deteriorates.
[0009]
Therefore, in the encoding device shown in FIG. 23, considering that the magnitude of the inter-frame difference signal of the input image affects the amount of generated codes of the inter-frame predictive coding, the encoding device shown in FIG. The low pass band of the prefilter for the input image is limited according to the size. As a result, the energy of the prediction error signal S212 is attenuated, and coarse quantization can be prevented, so that subjective image quality can be improved.
[0010]
A variable prefilter control method in the encoding device will be described.
[0011]
The computing unit 204 calculates an inter-frame difference r between the image signal S210 input from the terminal 201 and the image signal S214 input from the frame memory 212. Here, the image signal S214 is a signal referred to in the inter-frame prediction of the input image signal S210. The inter-frame difference r is input to the pre-filter controller 203, and the pre-filter controller 203 determines a pre-filter parameter as a parameter for controlling the low-pass passband of the pre-filter 202 according to the magnitude of the inter-frame difference r. Generate and output a filter coefficient k. The pre-filter 202 has the two-dimensional low-pass filter characteristic shown in FIG. 24, and the low-pass band is monotonically increased with respect to the pre-filter coefficient k. FIG. 25 shows the relationship between the inter-frame difference r and the pre-filter coefficient k. As described above, the pre-filter controller 203 controls the low-pass characteristic of the pre-filter according to the magnitude of the inter-frame difference signal of the input image according to the above relationship.
[0012]
[Problems to be solved by the invention]
By the way, in the frame, there are a portion where the prediction residual is large and a portion where the prediction residual is small locally. For example, the former is a locally fast moving part or a part with a complicated pattern, and the latter is a slowly moving part or a part with a flat pattern. The human eye is insensitive to a decrease in resolution in a fast-moving part or a part with a complicated picture, but is sensitive to a decrease in resolution in a slow-moving part or a part with a flat picture. Therefore, it is preferable to use a low-pass filter having a relatively narrow pass band in a portion having a large prediction residual, and to use a low-pass filter having a relatively wide pass band in a portion having a small prediction residual. The amount of generated information is reduced, and the subjective image quality is improved.
[0013]
However, as in the configuration of FIG. 23, since the pre-filter 202 is placed outside the predictive coding loop of the coding circuit 205, if the characteristics of the pre-filter are frequently changed partially inside the frame, If the passband of the low-pass differs between predicted images, the coding efficiency may be reduced. That is, if the bands of the image signals are different due to the difference in the characteristics of the pre-filters used between the predicted images, the energy of the input image signal S210 and the energy of the prediction residual signal of the predicted image signal S214 in the configuration of FIG. Also, there is a possibility that the energy of the prediction residual signal between the filtered image signal S211 and the predicted image signal S214 is larger. For example, when the input image signal S210 is low-pass filtered and the high frequency component is attenuated in the image signal S211 while the high frequency component remains in the predicted image signal S214, the input image signal S210 and the predicted image signal The energy of the prediction residual signal of the filtered image signal S211 and the prediction residual signal of the predicted image signal S214 becomes larger than the energy of the prediction residual signal of S214, and there is a possibility that the coding efficiency is reduced.
[0014]
Therefore, the present invention has been made in view of such a situation, and in a moving image coding apparatus having a configuration in which a pre-filter is provided in a stage preceding an inter-frame predictive coding circuit, a low-pass This makes it possible to change the characteristics of the pre-filter and to reduce a decrease in the coding efficiency due to the difference in the band of the image signal between the predicted images.
[0015]
That is, an object of the present invention is to provide a moving image encoding method and a moving image encoding method that can improve the subjective image quality by partially changing the characteristics of a low-pass prefilter in accordance with a difference between local prediction residuals in a frame. An image encoding device is provided.
[0016]
Another object of the present invention is to provide a moving image encoding method and a moving image encoding method capable of improving the subjective image quality by reducing the decrease in encoding efficiency due to the difference in the bandwidth of each image signal between images to be referred to. An image encoding device is provided.
[0017]
Another object of the present invention is to provide a moving picture coding method and a moving picture coding method for reducing the deterioration of subjective image quality due to the extremely different band of image signals between temporally adjacent pictures. To provide a chemical conversion device.
[0018]
Still another object of the present invention is to provide a moving picture coding method and a moving picture coding method that reduce subjective image quality deterioration due to extremely different image signal bands between spatially adjacent small blocks in the same image. An object of the present invention is to provide a moving picture encoding device.
[0019]
[Means for Solving the Problems]
According to the present invention, an input moving image is divided into small blocks each including at least one pixel, the encoding difficulty of the small block is calculated by a predetermined method, and the encoding difficulty of the small block and the small block are inter-picture predicted. The above-mentioned problem is solved by determining filter characteristics when adaptively performing low-pass filter processing on the small block based on the filter control information used for the small block referred to in (1).
[0020]
That is, according to the present invention, when determining the filter characteristics at the time of the low-pass filter processing on the input video signal, together with the coding difficulty of the current small block to be coded, the small block referred to in the inter-picture prediction is determined. The filter control information used in the block is also considered. This makes it possible for the moving image subjected to the low-pass pre-filter processing to have a better subjective impression of the coding image quality and higher coding efficiency than before.
[0021]
In the present invention, the input moving image is classified into a filter characteristic image to be referred to for determining the filter characteristic and a non-filter characteristic image not to be referred to, and the filter characteristic is determined as follows.
[0022]
That is, when the input moving image is a filter characteristic image, the input moving image is divided into small blocks each including at least one pixel, and the coding difficulty of the small block is calculated. Is smoothed. Then, based on the encoding difficulty of the small block and the filter control information used in the small block referred to in the inter-picture prediction, the small block is adaptively subjected to the low-pass filter processing. Determine the filter characteristics.
[0023]
When the input moving image is a non-filter characteristic image, the image is divided into small blocks each including at least one pixel, and the small blocks of the input image are referred to in the inter-picture prediction. A filter characteristic when low-pass filtering is performed on the filter characteristic of the small block is calculated by interpolation using a predetermined method based on the control information and the filter control information of the small block in the future reference image. Further, a final filter characteristic is determined based on the filter control information of the small blocks at the same spatial position in the adjacent images.
[0024]
As described above, according to the present invention, when determining the filter characteristics at the time of the low-pass filter processing on the input moving image signal, images having similar coding difficulty characteristics as the filter characteristic images, such as a P picture in the MPEG system, etc. , To calculate a filter characteristic, and select an image having a different encoding difficulty characteristic from the filter characteristic image as a non-filter characteristic image, for example, an I-picture or a B-picture in the MPEG system. By calculating the filter characteristic by performing interpolation, it is possible to prevent the low-pass band between successive images from being extremely different.
[0025]
In addition, in both the filter characteristic image and the non-filter characteristic image, by considering the filter control information used in the small block that the current encoding target small block refers to in the inter-picture prediction, an extreme difference in the filter coefficient is obtained. To prevent the coding efficiency from decreasing.
[0026]
As a result, the moving image subjected to the low-pass pre-filter processing can have a better subjective impression of the coding image quality and a higher coding efficiency than before.
[0027]
BEST MODE FOR CARRYING OUT THE INVENTION
Hereinafter, preferred embodiments of the present invention will be described with reference to the drawings.
[0028]
FIG. 1 shows a first configuration example of a moving picture coding apparatus to which the moving picture coding method of the present invention is applied.
[0029]
In the moving picture coding apparatus according to the first configuration example of the present invention shown in FIG. 1, a digital moving picture signal S30 input from an input terminal 20 is sent to a frame memory group 21 and stored.
[0030]
Using the image data S23 stored in the frame memory group 21, the motion vector detection circuit 24 detects a motion vector between frames. More specifically, the motion vector detection circuit 24 divides the frame into small blocks, and calculates a motion vector MV for each small block. Here, the small block is composed of, for example, 16 pixels × 16 lines, and the motion vector detection is performed, for example, by pattern matching between the reference frame and the current small block. That is, as shown in Expression (1), the sum Ef of the absolute value of the difference between the signal Aij of the current small block and the signal Fij of the small block referenced by an arbitrary motion vector is obtained.
[0031]
Ef = Σ | Aij−Fij | (i = 0 to 15, j = 0 to 15) (1)
The motion vector detection circuit 24 outputs a motion vector having the minimum value of Ef as a motion vector signal MV.
[0032]
The image data S23 and the motion vector signal MV are input to the encoding difficulty measuring circuit 26, where the encoding difficulty d (d_current) is calculated for each small block of the input image. The encoding difficulty d in the first configuration example is a parameter representing the difficulty when the amount of code generated in the moving image encoding circuit 23 described later is compressed to a predetermined bit rate. A specific configuration example of the encoding difficulty measuring circuit 26 will be described later.
[0033]
Next, the encoding difficulty d of the current small block to be encoded obtained as described above is input to the prefilter control circuit 27.
[0034]
The pre-filter control information storage circuit 25 stores pre-filter coefficients used in the small blocks input in the past. Here, a block address mb_address indicating the position of the current small block on the frame and the motion vector signal MV of the current small block are input to the pre-filter control information storage circuit 25. The pre-filter control information storage circuit 25 reads out a filter coefficient k_ref as a parameter for designating the pass band limitation of the low-pass pre-filter used in the small block referred to in the inter-frame prediction by the current small block. The filter coefficient k_ref is input to the pre-filter control circuit 27. When the filter coefficient k_ref is read, only the block address mb_address is used and the pass band of the low-pass pre-filter used in the small block located at the same block address mb_address on the frame referred to in the inter-frame prediction. A parameter designating the restriction may be read as the filter coefficient k_ref. In this case, there is no need to input the motion vector signal MV to the prefilter control information storage circuit 25.
[0035]
When the encoding difficulty d and the filter coefficient k_ref are input, the pre-filter control circuit 27 uses the filter coefficient k_current as a parameter to specify the pass band limitation of the low-pass pre-filter used for the current small block based on these. Is generated and input to the pre-filter 22. Here, the characteristics of the pre-filter 22 are, for example, the characteristics as shown in FIG. Therefore, here, as the filter coefficient k_current is smaller, the low-pass filter characteristic has a narrower pass band.
[0036]
The characteristics of the pre-filter 22 are not limited to the low-pass filter having the characteristics shown in FIG. 24, and may be low-pass filters having other characteristics.
[0037]
The filter coefficient k_current calculated by the prefilter control circuit 27 is also input to and stored in the prefilter control information storage circuit 25, and is used as a reference value for determining a prefilter coefficient to be used for a small block to be input in the future. (Filter coefficient k_ref).
[0038]
An example of a method for calculating the filter coefficient k_current will be described below. Here, as shown in FIG. 2, a group is formed for each N frames (N ≧ 1) in the input moving image input order, and this group is set as one unit of processing. Hereinafter, control of the pre-filter coefficient will be described with reference to the flowchart of FIG.
[0039]
In the flowchart shown in FIG. 3, first, in step ST40, control of a filter between N frames including the current encoding target frame is started.
[0040]
In step ST41, the average value d_ave of the encoding difficulty levels between the N frames is calculated, and in the next step ST42, the representative value k_gop of the pre-filter coefficient between the N frames is calculated. With respect to the representative value k_gop of the pre-filter coefficient, the correspondence to the representative (average) pre-filter coefficient with respect to the average value d_ave of the encoding difficulty between N frames is empirically determined in advance. . Here, as described with reference to FIG. 25, as the encoding difficulty increases, a low-pass filter characteristic having a narrower pass band is associated (that is, k_gop decreases). The correspondence between the average value d_ave of the encoding difficulty and the representative value k_gop of the pre-filter coefficient is as shown in FIG. 4, for example.
[0041]
Next, in step ST43, control of the filter within one frame is started.
[0042]
First, in step ST44, the encoding difficulty d (hereinafter, referred to as d_current) of the current small block to be encoded is read, and in step ST45, the current small block is referred to in inter-frame prediction as shown in FIG. The low-pass prefilter coefficient k_ref used in the small block is read.
[0043]
Next, in step ST46, a low-pass pre-filter coefficient k_current to be used in the current small block is calculated from the encoding difficulty d_current and the low-pass pre-filter coefficient k_ref.
[0044]
FIG. 5 shows a calculation example of the low-pass pre-filter coefficient k_current. In the calculation example shown in FIG. 5, k_current is first statistically calculated from d_current and k_gop as shown in Expression (a). Here, if d_current is larger than d_ave, k_current is calculated. On the other hand, if the value of d_current is smaller than the value of d_ave, the value of k_current is larger than the value of k_gop (ie, the characteristic of the low-pass filter has a wide pass band). ). The correspondence between the encoding difficulty d_current and the low-pass prefilter coefficient k_current is, for example, as shown in FIG.
[0045]
In the calculation example shown in FIG. 5, next, as in the condition (b) and the condition (c), the low-pass pre-filter coefficient k_current and the low-pass used in the small block referenced by the current small block in the inter-frame prediction are used. Compare with the pre-filter coefficient k_ref. When the change in k_current and k_ref is larger than a predetermined threshold, control is performed to suppress the change. Specifically, in the calculation example shown in FIG. 5, when k_current is larger than A times k_ref (A> 1), as in the condition (b), the k_current is changed to k_ref which is A times. Also, as in the condition (c), when the above k_current is smaller than B times (B <1) of the above k_ref, the k_current is changed to k_ref times B times. As described above, the low-pass pre-filter coefficient k_current used in the current small block is calculated.
[0046]
Returning to the flowchart of FIG. 3, in step ST47 following step ST46, it is determined whether or not the current small block is the last block in the frame. If it is determined in step ST47 that the current small block is not the last block in the frame, the process returns to step ST44. On the other hand, if it is determined in step ST47 that the current small block is the last block in the frame, the process proceeds to step ST48.
[0047]
In step ST48, it is determined whether or not the current frame is the last frame of the group of N frames. If it is determined in step ST48 that the current frame is not the last frame of the group including N frames, the process returns to step ST43. On the other hand, it is determined that the current frame is the last frame of the group including N frames. If so, the process proceeds to step ST49.
[0048]
In this step ST49, control of the filter in the group consisting of N frames is ended.
[0049]
As described above, the pre-filter control circuit 27 determines the pre-filter coefficient k_current.
[0050]
In the flowchart of FIG. 3, in steps ST45 and ST46, instead of using the filter coefficient k_ref, the encoding difficulty d_ref of the small block referred to in the inter-frame prediction by the current small block may be used.
[0051]
An example of a method for calculating the pre-filter coefficient k_current in this case will be described with reference to the flowchart shown in FIG. The flowchart of FIG. 7 differs from the flowchart of FIG. 3 in steps ST55 and ST56, which correspond to steps ST45 and ST46 in FIG. 3, respectively. Steps ST50 to ST54 in FIG. 7 are the same as steps ST40 to ST44 in FIG. 3, and steps ST57 and subsequent steps in FIG. 7 are the same as steps ST47 and subsequent steps in FIG. The description of the steps is omitted.
[0052]
In the flowchart of FIG. 7, in step ST55, as shown in FIG. 2, the encoding difficulty level d_ref of the small block referenced by the current small block in inter-frame prediction is read.
[0053]
Next, in step ST56, a low-pass pre-filter coefficient k_current used in the current small block is calculated from the coding difficulty d_current of the current small block to be coded and the coding difficulty d_ref.
[0054]
FIG. 8 shows a calculation example of the low-pass pre-filter coefficient k_current. In the calculation example shown in FIG. 8, as in the conditions (d) and (e), the coding difficulty d_current of the current small block to be coded and the current small block referred to in the inter-frame prediction by the current small block. Compare the encoding difficulty d_ref used in the block. When the change in d_current and d_ref is larger than a predetermined threshold, control is performed to suppress the change. Specifically, as in the condition (d), when d_current is larger than C times (C> 1) of d_ref, the d_current is changed to d_ref which is C times. Also, as in the condition (e), when d_current is smaller than D times (D <1) of d_ref, d_current is changed to d times d_ref. As described above, the encoding difficulty d_current used in the current small block is checked. Next, as shown in the equation (f) of FIG. 8, k_current is temporarily calculated statistically from the d_current and k_gop. The characteristics at this time are the same as those described with reference to FIG.
[0055]
In the case of the flowchart of FIG. 7 described above, the encoding difficulty d_ref of the small block calculated in the past is stored in the pre-filter control information storage circuit 25 of FIG. 1, and the block address mb_address of the current small block and The coding difficulty d_ref of the small block referred to in the inter-frame prediction is read from the motion vector signal MV.
[0056]
At this time, as described above, when reading the encoding difficulty level d_ref, only the block address mb_address is used, and the small block located at the same block address mb_address on the frame referred to in the inter-frame prediction is used. A parameter designating the pass band limitation of the used low-pass pre-filter may be read as the filter coefficient d_ref. In this case, there is no need to input the motion vector signal MV to the prefilter control information storage circuit 25.
[0057]
Returning to FIG. 1, the pre-filter 22 performs a low-pass filter process specified by the pre-filter coefficient k_current on the current small block, and outputs a processed image signal S22.
[0058]
The processed image signal S22 and its motion vector signal MV are input to a moving image coding circuit 23, where a predetermined inter-frame predictive coding process is performed, and output from an output terminal 28 as a coded bit stream S24. .
[0059]
FIG. 9 shows a configuration example of the moving picture coding circuit 23 of FIG. FIG. 9 shows a specific example of the moving picture coding circuit 23 that performs hybrid coding combining motion compensation inter-frame prediction and DCT as in the so-called MPEG method, for example.
[0060]
In FIG. 9, a motion vector signal MV input from the motion vector detection circuit 24 is supplied to an input terminal 61. On the other hand, the input terminal 50 supplies an input video signal S65 to the video coding circuit (hybrid encoder) 23.
[0061]
Further, the motion compensation inter / intra prediction circuit 57 of the encoding circuit 23 includes an image memory, and outputs a predicted image signal S68 read from the image memory based on the motion vector signal MV from the input terminal 61. .
[0062]
The arithmetic unit 51 performs an addition process by using the input moving image signal S65 from the input terminal 50 as an addition signal and the prediction image signal S68 from the motion compensation inter / intra prediction circuit 57 as a subtraction signal. The difference between the input moving image signal S65 and the prediction image signal S68 is calculated, and the difference is output as the prediction residual signal S66. When there is a scene change, prediction is not performed, and the input moving image signal S65 is taken out as it is.
[0063]
Next, the prediction residual signal S66 (or the original signal when prediction is not performed) is sent to the DCT circuit 52. The DCT circuit 52 performs two-dimensional DCT on the prediction residual signal S66. The DCT coefficients output from the DCT circuit 52 are scalar-quantized by the quantization circuit 53. The quantized output signal of the quantization circuit 53 is sent to a variable length coding (VLC) circuit 58 and an inverse quantization circuit 54. The VLC circuit 58 performs, for example, Huffman coding on the quantized output signal. The output signal of the VLC circuit 58 is sent to a buffer memory 59. The buffer memory 59 smoothes the bit rate of the data string output from the output terminal 60 to the transmission path. When the buffer memory 59 is about to overflow, the fact is fed back to the quantization circuit 53 as quantization control information. At this time, the quantization circuit 53 increases the quantization step, thereby reducing the amount of information output from the quantization circuit 53.
[0064]
On the other hand, the inverse quantization circuit 54 performs an inverse quantization process on the quantized output signal in accordance with the quantization step information q_step supplied from the quantization circuit 53. The output of the inverse quantization circuit 54 is input to an inverse DCT circuit 55, where the prediction residual signal S67 decoded by inverse DCT processing is input to an arithmetic unit 56.
[0065]
Further, the same signal as the predicted image signal S68 supplied to the calculator 51 is supplied to the calculator 56. The arithmetic unit 56 adds the prediction image signal S68 to the prediction residual signal S67. Thereby, a locally decoded image signal is obtained. This image signal is the same signal as the output image on the receiving side.
[0066]
Next, FIG. 10 shows a configuration example of the encoding difficulty measuring circuit 26 of FIG. The configuration of FIG. 10 is basically the same as that of the moving picture coding circuit described with reference to FIG. 9, except that a fixed quantization scale is used in the quantization circuit 73 and the VLC circuit 78. The point is that the occupation amount of the buffer memory is not managed for the amount of code generated from. That is, the bit amount of the generated code amount from the VLC circuit 78 is counted for each small block by the counter 79, and the encoding difficulty d is output from the output terminal 80. The description of the same components as those in FIG. 9 will be omitted.
[0067]
In the moving picture coding apparatus having the above-described configuration, the current frame to be coded of the input moving picture signal S30 is divided into small blocks each including at least one pixel, and coding is difficult in units of small blocks. Degree (a parameter indicating the degree of difficulty in compressing the generated code amount of the moving picture coding apparatus to a predetermined bit rate), and performs a low-pass pre-filter process on the current small block to be coded. In this case, adaptively determine the characteristics of the low-pass pre-filter based on the encoding difficulty of the current small block and the pre-filter control information used in the small block referred to in the inter-frame prediction by the current small block, Encoding processing is performed on the moving image signal that has been subjected to the low-pass pre-filter processing according to this characteristic.
[0068]
That is, in the video encoding device of the first configuration example of the present invention, when determining the filter characteristics at the time of the pre-filter process on the video signal to be subjected to the inter-frame predictive encoding, the small current encoding target is determined. The pre-filter control information used in the small block referred to in the inter-frame prediction is considered together with the coding difficulty of the block. Specifically, first, as a first step, a low-pass filter having a relatively narrow pass band is selected in a portion having a high degree of coding difficulty due to a locally fast motion or a complicated pattern in consideration of visual characteristics. On the other hand, a low-pass filter having a relatively wide pass band is selected in a portion having a low degree of difficulty in coding due to a slow motion or a flat picture. Next, as a second step, considering the pre-filter control information used in the small block in which the current encoding target small block is referred to in the inter-frame prediction, the inter-frame prediction encoding efficiency is not reduced. The filter characteristics selected in one stage are corrected so that the pre-filter characteristics for the current small block to be encoded are finally determined. As a result, the moving image subjected to the low-pass pre-filter processing has a good subjective impression of the coding image quality and has high coding efficiency.
[0069]
Next, FIG. 11 shows a moving image encoding apparatus according to a second configuration example of the present invention.
[0070]
The major difference between the second configuration example shown in FIG. 11 and the video encoding device of the first configuration example described above is that the encoding difficulty d (d_current) input to the pre-filter control circuit 97 is The prediction residual output from the vector detection circuit 96, that is, the sum of the absolute value of the difference between the signal Aij of the current block and the signal Fij of the block referenced by an arbitrary motion vector, which is calculated by the above equation (1). The point is that Ef is used.
[0071]
In FIG. 11, a digital video signal S100 input from an input terminal 90 is sent to a frame memory group 91 and stored. Using the image data S103 stored in the frame memory group 91, the motion vector detection circuit 96 detects a motion vector between frames as described in the first configuration example. Then, the motion vector detection circuit 96 calculates a prediction residual Ef calculated from the motion vector signal MV and the above equation (1) for each small block of the input image. The prediction residual Ef is output from the motion vector detection circuit 96 as the encoding difficulty d of the small block.
[0072]
The pre-filter control circuit 97 receives the current encoding difficulty d of the small block to be encoded. Further, the pre-filter control information storage circuit 95 stores pre-filter coefficients used in small blocks input in the past. Here, a block address mb_address indicating the position of the current small block on the frame and the motion vector signal MV of the current small block are input to the prefilter control information storage circuit 95. In step 95, the filter coefficient k_ref is read out as a parameter for designating the pass band limitation of the low-pass pre-filter used in the small block referred to in the inter-frame prediction by the current small block, input. When the filter coefficient k_ref is read, only the block address mb_address is used to limit the pass band of the low-pass pre-filter used in the small block located at the same block address mb_address on the frame referred to in the inter-frame prediction. The designated parameter may be read as the filter coefficient k_ref. In this case, there is no need to input the motion vector signal MV to the prefilter control information storage circuit 95.
[0073]
When the encoding difficulty d and the filter coefficient k_ref are input, the pre-filter control circuit 97 sets the filter coefficient k_current as a parameter for designating the pass band limitation of the low-pass pre-filter used for the current small block based on these. The output is generated and input to the pre-filter circuit 92. Here, the characteristics of the pre-filter circuit 97 are, for example, the characteristics as shown in FIG. Therefore, here, as the filter coefficient k_current is smaller, the low-pass filter characteristic has a narrower pass band.
[0074]
The characteristics of the pre-filter 92 are not limited to the low-pass filter having the characteristics shown in FIG. 24, and may be low-pass filters having other characteristics.
[0075]
The filter coefficient k_current calculated by the pre-filter control circuit 97 is also input to and stored in the pre-filter control information storage circuit 95, and is used as a reference value for determining a pre-filter coefficient to be used for a small block to be input in the future. It will be used as k_ref.
[0076]
An example of a method of calculating the filter coefficient k_current is the same as the method described with reference to FIG. 5 in the first configuration example.
[0077]
The pre-filter circuit 92 performs a low-pass filter process specified by the pre-filter coefficient k_current on the current small block, and outputs a processed image signal S102.
[0078]
The processed image signal S102 and its motion vector signal MV are input to a moving image encoding circuit 93, and are subjected to a predetermined inter-frame prediction encoding process (for example, a so-called MPEG encoding process). S104 is output from the output terminal 98. The configuration of the video encoding circuit 93 is the same as that described with reference to FIG. 9 in the first configuration example.
[0079]
In the moving picture coding apparatus having such a configuration, as the coding difficulty d (d_current) input to the prefilter control circuit 97, the prediction residual output from the motion vector detection circuit 96, that is, the signal of the current block, By using the sum Ef of the absolute value of the difference between the signal Aij and the signal Fij of the block referenced by an arbitrary motion vector, the selected filter characteristic is modified so that the inter-frame prediction coding efficiency is not reduced, and the final The pre-filter characteristic for the current small block to be encoded is determined. Thus, the moving image that has been subjected to the low-pass pre-filter processing can have a good subjective impression of the coding image quality and high coding efficiency.
[0080]
The pre-filter control method and apparatus in moving image coding disclosed in Japanese Patent Application Laid-Open No. 6-225276 also perform pre-filter processing on an input image signal. The pre-filter coefficient is not changed in one screen, but is constant in the screen, and the filter coefficient is constant in units of a plurality of frames.
[0081]
On the other hand, in the image coding apparatuses of the first configuration example and the second configuration example as described above, the filter characteristics at the time of the pre-filter processing on the moving image signal to be subjected to the inter-frame prediction coding are displayed on the screen. Since it is adaptively changeable in accordance with the encoding difficulty d of the small block unit to be divided, it is a local fast motion or a complicated pattern in the screen and has a large encoding difficulty in consideration of visual characteristics. In the part, a low-pass filter having a relatively narrow pass band can be selected. On the other hand, in the part having a slow motion or a flat pattern and having a low coding difficulty, a low-pass filter having a relatively wide pass band can be selected. Further, when determining the pre-filter coefficient in the small block unit, not only the coding difficulty d of the small block but also the current small block to be coded is used in the small block referred to in the inter-frame prediction. The pre-filter control information is also taken into account, and the filter characteristics selected as described above are modified so that the inter-frame prediction Since the pre-filter characteristic is determined for the moving image, the moving image subjected to the low-pass pre-filter processing has a better subjective impression of the coding image quality and higher coding efficiency than the conventional one.
[0082]
By the way, in the inter-frame prediction coding used in the MPEG system or the like, as shown in FIG. 12, an I picture to be internally coded without performing prediction, a B picture to perform only forward prediction, and a bi-directional prediction are used. There are B pictures to be performed, and in general, the dynamic range of the encoding difficulty and the distribution of the encoding difficulty for each small block differ between these images. In the MPEG method, since images having different characteristics of encoding difficulty are continuous as described above, if the filter characteristic of the low-pass filter processing is determined according to the encoding difficulty of all the input images, the temporally adjacent images are determined. There is a possibility that subjective image quality will be degraded due to extremely different filter characteristics.
[0083]
In the embodiment described using the first and second configuration examples, the encoding difficulty of the small block of the input image and the filter control information used in the small block that the small block refers to in the inter-picture prediction are described. , The filter characteristics of the low-pass filter process are determined based on the I-picture and the P-picture in which the B-picture exists between the I-picture to be subjected to the internal coding and the reference image having the filter characteristics, and the temporally consecutive pictures. No consideration is given to the filter characteristics.
[0084]
Also, in an image for which inter-frame prediction is performed, there may be a small block for which intra-coding is performed without performing prediction. Therefore, even in the same image, it is extremely small compared to a spatially adjacent small block. There may be small blocks with high coding difficulty.
[0085]
When the filter characteristics of the low-pass filter processing are adaptively determined for such an image in accordance with the encoding difficulty of each small block, there are small blocks having different low-pass pass bands locally in time / space. As a result, the subjective image quality may be degraded.
[0086]
Therefore, in the moving picture coding apparatus according to the third configuration example of the present invention shown in FIG. 13, when determining the filter characteristics at the time of pre-filtering a moving picture signal to be subjected to inter-frame predictive coding, the input moving picture The current encoding target frame of the signal S30 is divided into small blocks each including at least one pixel, and it is determined whether the current encoding target image is a filter characteristic image or a non-filter characteristic image. The pre-filter information of the current small block is used together with the pre-filter control information used in the small block referred to in the inter-frame prediction and the pre-filter used in the small block at the same spatial position in the temporally adjacent image. The filter control information is also considered.
[0087]
Specifically, first, as a first step, an input image of a specific coding type is filtered by taking into consideration that the coding difficulty varies depending on the coding type such as I, P, and B pictures in the MPEG system. The image is determined as an image, and the other coding type images are determined as non-filter characteristic images. Here, for example, a P picture is used for the filter characteristic image, and I and B pictures are used for the non-filter characteristic image.
[0088]
When the input image is determined as the filter characteristic image, the encoding difficulty for each small block of the image is calculated by a predetermined method, and the encoding difficulty is smoothed in the image by a predetermined method.
[0089]
As a second step, in consideration of visual characteristics, a low-pass filter having a relatively narrow pass band is selected for a small block having a large coding difficulty due to a local fast motion or a complicated picture, while a slow motion is selected. For a small block having a small coding difficulty due to a flat pattern or a flat pattern, a low-pass filter having a relatively wide pass band is selected. Furthermore, in consideration of the pre-filter control information used in the small block that the current encoding target small block refers to in the inter-frame prediction, the image quality is not subjectively degraded, and the inter-frame prediction encoding efficiency is reduced. The pre-filter characteristics for the current small block to be encoded are determined by modifying the selected filter characteristics so as not to perform the processing.
[0090]
In the case of a non-filter characteristic image, as the second stage, the pre-filter control information used in the small block of the past reference image and the small block of the future reference image used by the current small block to be encoded are referred to in inter-frame prediction. Interpolation is performed by a predetermined method from the obtained prefilter control information to calculate prefilter characteristics of a small block of the image. Furthermore, taking into account the control information of the pre-filter used in the small blocks at the same spatial position in the temporally close image, the image quality is not subjectively degraded and the inter-frame predictive coding efficiency is lowered. The calculated filter characteristic is corrected so that the pre-filter characteristic for the current small block to be encoded is determined.
[0091]
Here, when the non-filter characteristic image is an intra-coded image that does not use prediction, for example, an I-picture in the MPEG system, the pre-filter control information of a small block at the same spatial position in a P picture that precedes and follows in time. To determine the pre-filter characteristics of the current small block to be encoded.
[0092]
As a result, the moving image that has been subjected to the low-pass pre-filter processing has a good subjective impression of the coding image quality and has high coding efficiency.
[0093]
In the moving picture coding apparatus according to the third configuration example of the present invention shown in FIG. 13, the digital moving picture signal S30 input from the input terminal 120 is sent to the frame memory group 121 and stored.
[0094]
In the MPEG system, bidirectional prediction is performed when predictive coding is performed. Therefore, as shown in FIGS. 14A and 14B, the image input order and the coding order are different. Then, the order of the input images is changed according to the coding order.
[0095]
Using the image data S23 stored in the frame memory group 121, the motion vector detection circuit 124 detects a motion vector between frames. More specifically, the motion vector detection circuit 124 divides the frame into small blocks, and calculates the motion vector MV for each small block. Here, the small block is composed of, for example, 16 pixels × 16 lines, and the motion vector detection is performed, for example, by pattern matching between the reference frame and the current small block. That is, the sum Ef of the absolute value of the difference between the signal Aij of the current small block and the signal Fij of the small block referred to by an arbitrary motion vector is obtained by the above equation (1).
[0096]
The motion vector detection circuit 124 outputs a motion vector having the minimum value of Ef as a motion vector signal MV.
[0097]
The image processing type determination circuit 128 changes the contents of the output circuit switching signal pict_type according to the picture type of the input image data S23, and changes the subsequent processing.
[0098]
If the image data S23 is an image used as a filter characteristic image, for example, a P picture, the image data S23 is input to the encoding difficulty level reference circuit 126, and the circuit switching signal pict_type (1) is input to the motion vector detection circuit 124 and This is input to the pre-filter control information storage circuit 125.
[0099]
In the motion vector detection circuit 124 to which the circuit switching signal pict_type (1) is input, the motion vector signal MV is input to the video coding circuit 123, the prefilter control information storage circuit 125, and the coding difficulty measuring circuit 126.
[0100]
The encoding difficulty measuring circuit 126 to which the image data S23 and the motion vector signal MV are input calculates the encoding difficulty d (d_raw) for each small block of the input image.
[0101]
The encoding difficulty d (d_raw) in the third configuration example is a parameter indicating the difficulty when the amount of code generated in the moving image encoding circuit 123 described below is compressed to a predetermined bit rate. A specific configuration example of the encoding difficulty measuring circuit 126 will be described later.
[0102]
The encoding difficulty level smoothing circuit 129 smoothes the input d (d_raw) in a frame to avoid extremely different filter coefficients between adjacent small blocks, and obtains smoothed encoding difficulty level information d. (D_current) is calculated. In the encoding difficulty smoothing circuit 129, when the encoding difficulty of the i-th current small block to be encoded in the horizontal direction and j-th in the vertical direction is d (i, j), for example, (2)
d (d_current) = ｛d (i−1, j−1) + d (i, j−1) + d (i−1, j + 1) + d (i−1, j) + d (i, j) + d (i + 1, j) + d (i−1, j + 1) + d (i, j + 1) + d (i + 1, j + 1)｝ / 9 (2)
To perform smoothing.
[0103]
Next, the encoding difficulty level d (d_current) of the current encoding target small block obtained as described above is input to the prefilter control circuit 127.
[0104]
Further, the pre-filter control information storage circuit 125 stores pre-filter coefficients used in small blocks input in the past.
[0105]
Here, the pre-filter control information storage circuit 125 receives the block address mb_address representing the position of the current small block on the frame, the motion vector signal MV of the current small block, and the circuit switching signal pict_type (1). Based on these, in the pre-filter control information storage circuit 125, the filter coefficient k_ref () as a parameter for specifying the pass band limitation of the low-pass pre-filter used in the small block referred to in the inter-frame prediction by the current small block is 1) is read and input to the pre-filter control circuit 127.
[0106]
When reading out the filter coefficient k_ref (1) of the small block to be referred to in the inter-frame prediction, only the block address mb_address is used and the block is located at the same block address mb_address on the frame to be referred to in the inter-frame prediction. A parameter designating the pass band limitation of the low-pass pre-filter used in the small block may be read as the filter coefficient k_ref (1). In this case, there is no need to input the motion vector signal MV to the prefilter control information storage circuit 125.
[0107]
When the encoding difficulty d (d_current) and the filter coefficient k_ref (1) are input, the pre-filter control circuit 127 specifies the pass band limitation of the low-pass pre-filter used for the current small block based on these. A filter coefficient k_current is generated and output as a parameter to be input, and is input to the prefilter 122.
[0108]
The filter coefficient k_current calculated by the pre-filter control circuit 127 is also input to and stored in the pre-filter control information storage circuit 25, and is used as a reference value for determining a pre-filter coefficient to be used for a small block to be input in the future. (Filter coefficient k_ref).
[0109]
Next, if the image data S23 is a non-filter characteristic image, for example, an I or B picture in the MPEG system, the image processing type determination circuit 128 sets the block address mb_address indicating the current position of the small block on the frame. The circuit switching signal pict_type (2) is input to the prefilter control information storage circuit 125, and is input to the motion vector detection circuit 124 and the prefilter control information storage circuit 125.
[0110]
In the motion vector detection circuit 124 to which the circuit switching signal pict_type (2) has been input, the motion vector signal MV is input to the video encoding circuit 123 and the pre-filter control information storage circuit 125.
[0111]
The pre-filter control information storage circuit 125 to which the block address mb_address of the current small block and the motion vector MV of the current small block have been input is a small block of a future reference image and a past reference image which the current small block refers to in inter-frame prediction. Filter coefficients k_ref (2) and k_ref (3) as parameters for designating the pass band limitation of the low-pass pre-filter used in the low-pass pre-filter used in the small block at the spatially same position of the image temporally close to the filter coefficients k_ref (2) and k_ref (3) The filter coefficient k_ref (4) as a parameter for designating the pass band limitation of the filter is read and input to the interpolation circuit 130.
[0112]
When reading out the filter coefficients k_ref (2) and k_ref (3), only the block address mb_address is used and used in the small block located at the same block address mb_address on the frame referred to in the inter-frame prediction. The parameters for specifying the passband limitation of the low-pass prefilter may be read as the filter coefficients k_ref (2) and k_ref (3). In this case, there is no need to input the motion vector signal MV to the prefilter control information storage circuit 125.
[0113]
The interpolation interpolating circuit 130 to which the filter coefficients k_ref (2), k_ref (3), and k_ref (4) as the parameters of the current small block are input, sets the filter coefficient k_ref (2) of the past reference image and the future reference image. Interpolation is performed based on the filter coefficient k_ref (3), and the filter coefficient k_current of the current small block is calculated using the filter coefficient k_ref (4) of the small block at the same spatial position in the temporally adjacent image. Generate and output, and input this to the pre-filter 122.
[0114]
Here, the current small block is referred to in the inter-frame prediction without using the filter coefficient k_ref (4) used in the small block at the same spatial position in the image temporally adjacent to the current small block. It is also possible to read out only the filter coefficients k_ref (2) and k_ref (3) used in and to generate and output the filter coefficient k_current of the current small block.
[0115]
If the non-filter characteristic image is an image for which inter-frame prediction is not performed, such as an MPEG I-picture, the filter of a small block at the same spatial position in the temporally preceding and following filter characteristic image is used. The coefficients are read as the filter coefficient k_ref (2) of the past reference image and the filter coefficient k_ref (3) of the future reference image, and interpolation is performed. In this case, the filter coefficient k_ref (4) of the small block at the same spatial position in the temporally adjacent image is not used.
[0116]
The filter coefficient k_current calculated by the interpolation circuit is also input to and stored in the prefilter control information storage circuit 125, and is used as a reference value for determining a prefilter coefficient to be used for a small block to be input in the future ( It will be used as the filter coefficient k_ref).
[0117]
As described above, the filter coefficients of the pre-filter 122 are adaptively determined, and a coding scheme with a good subjective impression of the coding image quality and high coding efficiency is realized.
[0118]
Here, the characteristics of the pre-filter 122 are, for example, the characteristics shown in FIG. Therefore, here, as the filter coefficient k_current is smaller, the low-pass filter characteristic has a narrower pass band. As the pre-filter 122, a low-pass filter having the above-described characteristics shown in FIG. 24 or a low-pass filter having other characteristics can be used.
[0119]
An example of a method for calculating the filter coefficient k_current will be described below. Here, as shown in FIG. 15, a group is formed for each N frames (N ≧ 1) including both the filter characteristic image and the non-filter characteristic image in the input order of the input moving images, and this group is set as one unit of processing. . Control of the pre-filter coefficient will be described with reference to the flowchart in FIG.
[0120]
In the flowchart of FIG. 16, first, in step ST140, control of a filter among N frames including the current frame to be encoded is started.
[0121]
In step ST141, the average value d_ave of the encoding difficulty of the filter characteristic image between the N frames is calculated, and in the next step ST42, the representative value k_gop of the pre-filter coefficient between the N frames is calculated. As for the representative value k_gop of the pre-filter coefficient, the correspondence to the representative (average) pre-filter coefficient is previously determined empirically with respect to the average value d_ave of the encoding difficulty between N frames. . Here, as shown in FIG. 25 described above, as the encoding difficulty increases, a low-pass filter characteristic having a narrower pass band is associated (that is, k_gop decreases). The correspondence between the average value d_ave of the encoding difficulty and the representative value k_gop of the pre-filter coefficients is as shown in FIG. 6 described above, for example.
[0122]
Next, in step ST143, it is determined whether the current encoding target frame is a filter characteristic image or a non-filter characteristic image. If it is a filter characteristic image, the process proceeds to step ST144. If it is a non-filter characteristic image, the process proceeds to step ST151.
[0123]
In step ST144, control of the filter within one frame of the filter characteristic image is started. First, in step ST145, the smoothed encoding difficulty d (hereinafter, referred to as d_current) of the current small block to be encoded is read. In step ST146, the current small block is read as shown in FIG. The low-pass prefilter coefficient k_ref (1) used in the small block referred to in the inter-frame prediction is read.
[0124]
Next, in step ST147, the low-pass pre-filter coefficient k_current to be used in the current small block is calculated from the encoding difficulty d_current and the low-pass pre-filter coefficient k_ref (1), for example, in the same manner as the calculation example shown in FIG. I do.
[0125]
In the next step ST148, it is determined whether or not the current small block is the last block in the frame. If it is determined in step ST148 that the current small block is not the last block in the frame, the process returns to step ST145. On the other hand, if it is determined in step ST148 that the current small block is the last block in the frame, the process proceeds to step ST149.
[0126]
In step ST149, it is determined whether or not the current frame is the last filter characteristic image of the group including N frames. If it is determined in step ST149 that the current frame is not the last filter characteristic image of the group including N frames, the process returns to step ST143. On the other hand, the last filter characteristic image of the group including the current frame including N frames is determined. If it is determined that the process is performed, the process proceeds to step ST150.
[0127]
In step ST151, control of a filter within one frame of the non-filter characteristic image is started. First, in step ST152, as shown in FIG. 15B, the low-pass pre-filter coefficient k_ref (2) used in each of the small blocks of the past reference image and the future reference image, which the current small block refers to in the inter-frame prediction. And k_ref (3) are read, and in step ST153, the low-pass pre-filter coefficient k_ref (4) used in the small block at the same spatial position in the temporally adjacent frame is read as shown in FIG. Next, in step ST54, a low-pass pre-filter coefficient k_current used in the current small block is calculated from the filter coefficients k_ref (2), k_ref (3) and k_ref (4).
[0128]
FIG. 17 shows a calculation example of the low-pass pre-filter coefficient k_current. In the case where the current encoding target image, the past reference image, and the future reference image have a positional relationship as shown in FIG. 15B, in the calculation example shown in FIG. From (2) and k_ref (3), k_current is calculated by interpolation according to the distance of each image. Next, as in the conditions (h) and (i), the low-pass pre-filter coefficient k_ref used in the small block at the same spatial position in the image temporally adjacent to the small block in the current image to be encoded. Compare with (4). When the change in k_current and k_ref (4) is larger than a predetermined threshold, control is performed to suppress the change.
[0129]
Specifically, in the calculation example shown in FIG. 17, when the above k_current is larger than E times k_ref (4) (E> 1) as in the condition (h), the k_current is multiplied by E times k_ref (4). ). Further, as in the condition (i), when the k_current is smaller than F times (D <1) of k_ref (4), the k_current is changed to k_ref (4) which is F times. As described above, the low-pass pre-filter coefficient k_current used in the current small block is determined.
[0130]
When the non-filter characteristic image is an I picture, the low-pass filter coefficient used in the small block at the same position in the past reference image is set to k_ref (2), and the low-pass filter used in the small block at the same position in the future reference image is used. The same calculation is performed with the coefficient set as k_ref (3).
[0131]
Returning to the flowchart of FIG. 16, in step ST155 following step ST154, it is determined whether or not the current small block is the last block in the frame. If it is determined in step ST155 that the current small block is not the last block in the frame, the process returns to step ST152. On the other hand, if it is determined in step ST155 that the current small block is the last block in the frame, the process proceeds to step ST156.
[0132]
In step ST156, it is determined whether or not the current frame is the last non-filter characteristic image of the group including N frames. If it is determined in step ST156 that the current frame is not the last non-filter characteristic image of the group including N frames, the process returns to step ST156. On the other hand, the last non-filter characteristic of the group including the current frame including N frames is determined. If it is determined that the image is a characteristic image, the process proceeds to step ST150.
[0133]
In step ST150, the control of the filter in the group including N frames is ended.
[0134]
As described above, the prefilter control circuit 127 and the interpolation circuit 130 determine the prefilter coefficient k_current.
[0135]
In steps ST146, ST147, ST152, ST153, and ST154 in FIG. 16, the current small block is replaced with the filter coefficient k_ref (1), k_ref (2), k_ref (3), and k_ref (4). The coding difficulty d_ref (1), d_ref (2), d_ref (3), d_ref (4) of the small block referred to in the inter-frame prediction may be used.
[0136]
In this case, even when the image processing type determination circuit 128 determines that the input image is a non-filter characteristic image, the image data S123 is input to the encoding difficulty level measurement circuit 126, and the motion vector signal MV is output from the motion vector detection circuit 124. Is input to the encoding difficulty measuring circuit 126 to calculate the encoding difficulty d_ref (4).
[0137]
An example of a method of calculating the pre-filter coefficient k_current in this case will be described with reference to FIG. 18 differs from the flowchart of FIG. 16 described above in steps ST164, ST165, ST170, ST171, and ST172, and respectively corresponds to steps ST145, ST147, ST152, ST153, and ST154 in FIG. Steps ST158 to ST163, steps ST166 to ST169, and steps ST172 and after in FIG. 18 are the same as steps ST140 to ST145, steps ST148 to ST151, and steps after step ST155 in FIG. 16, respectively. Is omitted.
[0138]
In FIG. 18, in step ST164, as shown in FIG. 15A, the coding difficulty d_ref (1) of the small block referenced by the current small block in the inter-frame prediction is read. Next, in step ST165, the encoding difficulty d_current of the current small block to be encoded and the low-pass pre-filter coefficient k_current used in the current small block from the encoding difficulty d_ref (1) are shown in FIG. The calculation is performed in the same manner as the calculation example shown.
[0139]
In step ST70, the encoding difficulty d_ref (2) used in the small blocks of the past reference image and the future reference image, which the current small block refers to in the inter-frame prediction, as shown in FIG. And d_ref (3) are read, and in step ST53, as shown in FIG. 15 (b), the low-pass pre-filter coefficient k_ref (4) used in the small block at the same spatial position in the temporally adjacent frame is read. .
[0140]
FIG. 19 shows a calculation example of the low-pass pre-filter coefficient k_current. When the current encoding target image, the past reference image, and the future reference image have a positional relationship as shown in FIG. 15B, in the calculation example shown in FIG. 19, first, as shown in the equation (j), d_current is calculated from d_ref (2) and d_ref (3) by interpolation. Next, as in the conditions (k) and (l), the encoding difficulty level coefficient d_ref (4) of the small block at the same spatial position in the image temporally adjacent to the small block of the current encoding target image. ) And compare. When the change in d_current and d_ref (4) is larger than a predetermined threshold, control is performed to suppress the change.
[0141]
Specifically, in the calculation example shown in FIG. 19, when the above d_current is larger than C times (C> 1) of d_ref (4) as in the condition (k), the d_current is increased by d times C_d_ref (4). ). Further, as in the condition (1), when d_current is smaller than D times (D <1) of d_ref (4), d_current is changed to d_ref (4) which is D times. As described above, the encoding difficulty level d_current of the current small block is determined. Then, as in the condition (j), a low-pass pre-filter coefficient k_current used in the current small block is calculated from the above-mentioned encoding difficulty d_current and k_gop.
[0142]
When the non-filter characteristic image is an I picture, the coding difficulty of the small block at the same position in the past reference image is set to d_ref (2), and the coding difficulty of the small block at the same position in the future reference image is set to d_ref (2). Let d_ref (3).
[0143]
In the case of the flowchart of FIG. 18 described above, the encoding difficulty d_ref of the small block calculated in the past is stored in the pre-filter control information storage circuit 125, and the block address mb_address of the current small block and the motion vector signal The coding difficulty d_ref of the small block referred to in the inter-frame prediction is read from the MV.
[0144]
At this time, as described above, when reading the encoding difficulty level d_ref, only the block address mb_address is used, and the small block located at the same block address mb_address on the frame referred to in the inter-frame prediction is used. A parameter designating the pass band limitation of the used low-pass pre-filter may be read as the filter coefficient d_ref. In this case, there is no need to input the motion vector signal MV to the prefilter control information storage circuit 125.
[0145]
Returning to FIG. 13, the pre-filter 122 performs a low-pass filter process specified by the pre-filter coefficient k_current on the current small block, and outputs a processed image signal S122.
[0146]
The processed image signal S22 and its motion vector signal MV are input to the moving image coding circuit 123, where they are subjected to a predetermined inter-frame prediction coding process, and output from the output terminal 128 as a coded bit stream S24. .
[0147]
FIG. 20 shows a configuration example of the pre-filter 122. In the configuration example shown in FIG. 20, an input terminal 300 is supplied with an input image signal S300 to the pre-filter 122. On the other hand, the input terminal 305 receives the low-pass pre-filter coefficient k_current input from the pre-filter control circuit 127 or the pre-filter control circuit 130 by interpolation.
[0148]
Here, the low-pass pre-filter coefficient k_current is determined to take a value of 0 ≦ k_current ≦ 1.
[0149]
The input image signal S300 is input to the low-pass filter 301, where it is subjected to low-pass filtering. The processed image signal S301 by the low-pass filter 301 is input to the arithmetic unit 302. The arithmetic unit 302 outputs a difference signal S302 between the processed image signal S301 and the input image signal S300. here,
S302 = S301-S300
It is.
[0150]
Next, the difference signal S302 is input to the arithmetic unit 303. The arithmetic unit 303 outputs a signal S303 obtained by multiplying the difference signal S302 by a low-pass prefilter coefficient k_current. here,
S303 = S302 × k_current
It is.
[0151]
Next, the signal S303 from the arithmetic unit 303 is input to the arithmetic unit 304. The arithmetic unit 304 outputs a prefiltered image signal S304 which is an addition signal of the signal S303 and the input image signal S300. here,

It is.
[0152]
The moving image encoding circuit 123 has, for example, the configuration shown in FIG. 9 described above. The encoding difficulty measuring circuit 126 has, for example, the configuration shown in FIG.
[0153]
Next, a moving picture coding apparatus according to a fourth configuration example of the present invention is shown in FIG.
[0154]
In the fourth configuration example shown in FIG. 21, a great difference from the moving image encoding apparatus of the third configuration example is that the encoding difficulty d (d_current) input to the pre-filter control circuit 197 is The prediction residual output from the vector detection circuit 196, that is, the sum of the absolute value of the difference between the signal Aij of the current block and the signal Fij of the block referred to by an arbitrary motion vector, which is calculated by the above equation (1). The point is that Ef is used.
[0155]
In FIG. 21, a digital video signal S100 input from an input terminal 190 is sent to a frame memory group 191 and stored. Using the image data S103 stored in the frame memory group 191, the motion vector detection circuit 196 detects a motion vector between frames as described in the third configuration example.
[0156]
The motion vector detection circuit 196 calculates a motion vector signal MV and a prediction residual Ef calculated from the above equation (1) for each small block of the input image, and converts the motion vector signal MV into a pre- It is input to the filter control information storage circuit 195.
[0157]
The image processing type determination circuit 194 determines whether the image currently being processed is a filter characteristic image or a non-filter characteristic image based on the processing number pict_number of the input image data. If the image currently being processed is a filter characteristic image, for example, a P picture, the circuit switching signal pict_type (1) is input to the filter control information storage circuit 195 and the motion vector detection circuit 196.
[0158]
The motion vector detection circuit 96 to which the circuit switching signal pict_type (1) has been input outputs the prediction residual Ef as the encoding difficulty d (d_raw) of the small block.
[0159]
The encoding difficulty level smoothing circuit 109 smoothes the input d (d_raw) in the frame to avoid extremely different filter coefficients between adjacent small blocks, and obtains smoothed encoding difficulty level information d. Calculate (d_current). The smoothing process in the encoding difficulty smoothing circuit 199 is performed in the same manner as in the third configuration example. The encoding difficulty d (d_current) of the current small block to be encoded is input to the prefilter control circuit 197. The prefilter control information storage circuit 195 stores the prefilter coefficients used in the small blocks input in the past. Here, the pre-filter control information storage circuit 195 receives the block address mb_address indicating the position of the current small block on the frame, the motion vector signal MV of the current small block, and the circuit switching signal pict_type (1). Is done.
[0160]
The prefilter control information storage circuit 195 stores the filter coefficient k_ref (1) as a parameter for designating the passband limitation of the low-pass prefilter used in the small block referred to in the inter-frame prediction based on the current small block. Read and input this to the pre-filter control circuit 197.
[0161]
When the filter coefficient k_ref (1) is read, only the block address mb_address is used to pass through the low-pass pre-filter used in the small block located at the same block address mb_address on the frame referred to in the inter-frame prediction. A parameter for specifying the band limitation may be read as the filter coefficient k_ref. In this case, there is no need to input the motion vector signal MV to the prefilter control information storage circuit 195.
[0162]
When the encoding difficulty d and the filter coefficient k_ref (1) are input, the pre-filter control circuit 197 uses the filter as a parameter for designating the pass band limitation of the low-pass pre-filter used for the current small block based on these. A coefficient k_current is generated and output, and this is input to the prefilter 192. Here, the characteristics of the pre-filter 197 are, for example, the characteristics shown in FIG. Therefore, here, as the filter coefficient k_current is smaller, the low-pass filter characteristic has a narrower pass band.
[0163]
The characteristics of the pre-filter 192 are not limited to the low-pass filter having the characteristics shown in FIG. 24, and may be low-pass filters having other characteristics.
[0164]
The filter coefficient k_current calculated by the pre-filter control circuit 197 is input to and stored in the pre-filter control information storage circuit 195, and is used as a reference value for determining a pre-filter coefficient to be used for a small block to be input in the future. It will be used as k_ref.
[0165]
If the image processing type determination circuit 194 determines from the input pict_number that the image currently being processed is a non-filter characteristic image, for example, an I or B picture, the circuit switching signal pict_type (2) is filtered. The information is input to the information storage circuit 195 and the motion vector detection circuit 196, and the block address mb_address indicating the current position of the small block on the frame is input to the prefilter control information storage circuit 195 and the interpolation circuit 200.
[0166]
The motion vector detection circuit 196 to which the circuit switching signal pict_type (2) is input does not output the prediction residual Ef.
[0167]
The pre-filter control information storage circuit 195, to which the block address mb_address of the current small block, the motion vector MV of the current small block, and the circuit switching signal pict_type (2) are input, specifies the future in which the current small block refers in inter-frame prediction. Filter coefficients k_ref (2) and k_ref (3) as parameters for specifying the pass band limitation of the low-pass pre-filter used in the small block of the reference image and the past reference image, and spatially the same positions of the images temporally close to each other The filter coefficient k_ref (4) as a parameter for specifying the pass band limitation of the low-pass pre-filter used in the small block is read out and input to the interpolation circuit 200.
[0168]
When reading out the filter coefficients k_ref (2) and k_ref (3), only the block address mb_address is used and used in the small block located at the same block address mb_address on the frame referred to in the inter-frame prediction. The parameters for specifying the passband limitation of the low-pass prefilter may be read out as the filter coefficients k_ref (2) and k_ref (3). In this case, there is no need to input the motion vector signal MV to the prefilter control information storage circuit 195.
[0169]
The interpolation circuit 200 to which the motion vector signal MV of the current small block and the filter coefficients k_ref (2), k_ref (3), and k_ref (4) as the parameters are input, sets the filter coefficient k_ref (2) of the past reference image. ) And the filter coefficient k_ref (3) of the future reference image, perform interpolation, and further use the filter coefficient k_ref (4) of the small block at the same spatial position in the temporally adjacent image to determine the current small size. A filter coefficient k_current of the block is generated and output, and this is input to the pre-filter 292.
[0170]
Here, the current small block is referred to in the inter-frame prediction without using the filter coefficient k_ref (4) used in the small block at the same spatial position in the image temporally adjacent to the current small block. It is also possible to read out only the filter coefficients k_ref (2) and k_ref (3) used in and to generate and output the filter coefficient k_current of the current small block.
[0171]
If the non-filter characteristic image is an image for which inter-frame prediction is not performed, such as an MPEG I-picture, the filter of a small block at the same spatial position in the temporally preceding and following filter characteristic image is used. The coefficients are read as the filter coefficient k_ref (2) of the past reference image and the filter coefficient k_ref (3) of the future reference image, and interpolation is performed. In this case, the filter coefficient k_ref (4) of the small block at the same spatial position in the temporally adjacent image is not used.
[0172]
Also, when motion prediction is performed on an I picture, the I picture may be used as a filter characteristic image. Here, the characteristics of the pre-filter 192 are, for example, the characteristics shown in FIG. Therefore, here, as the filter coefficient k_current is smaller, the low-pass filter characteristic has a narrower pass band.
[0173]
The filter coefficient k_current calculated by the interpolation circuit is also input to and stored in the prefilter control information storage circuit 195, and is used as a reference value for determining a prefilter coefficient to be used for a small block to be input in the future ( It will be used as the filter coefficient k_ref).
[0174]
As described above, the filter coefficient of the pre-filter 192 is determined adaptively, and a coding scheme with a good subjective impression of the coding image quality and high coding efficiency is realized.
[0175]
The calculation method of the filter coefficient k_current is the same as the calculation example described with reference to FIG. 5 in the first configuration example.
[0176]
The prefilter 192 performs a low-pass filter process specified by the prefilter coefficient k_current on the current small block, and outputs a processed image signal S102.
[0177]
The processed image signal S102 and its motion vector signal MV are input to a moving image encoding circuit 193, and are subjected to a predetermined inter-frame predictive encoding process (for example, a so-called MPEG encoding process). S104 is output from the output terminal 198. The configuration of the video encoding circuit 193 is the same as that described with reference to FIG. 9 in the second configuration example.
[0178]
In the moving picture coding apparatus of the third configuration example, the filter characteristics at the time of prefiltering a moving picture signal to be subjected to inter-frame predictive coding are determined by encoding difficulty d in small block units for dividing a screen. Can be changed adaptively according to the visual characteristics.In consideration of the visual characteristics, a low-pass filter with a relatively narrow passband is used for parts that are locally fast moving in the screen or have complex patterns and are difficult to encode. On the other hand, a low-pass filter having a relatively wide pass band can be selected in a portion where the degree of coding difficulty is low due to slow motion or a flat picture.
[0179]
In addition, in consideration of an encoding method having a different encoding difficulty for each image such as the MPEG method, refer to the encoding difficulty in order to prevent the pre-filter coefficients from being extremely different between consecutive images. Image to be specified.
[0180]
Furthermore, in the moving picture coding apparatus of the fourth configuration example, not only the coding difficulty d of the small block but also the small block to which the current coding target small block refers in the inter-frame prediction and the neighboring image The pre-filter control information used in the small block at the same spatial position is also taken into account, and the filter characteristics selected as described above are modified so that the inter-frame prediction coding efficiency does not decrease. Finally, since the pre-filter characteristic for the current small block to be coded is finally determined, the low-pass pre-filtered moving image has a better subjective impression of the coded image quality than before, and In addition, coding efficiency can be improved.
[0181]
Here, the coded bit stream obtained by the coding in the moving picture coding device of the first to fourth configuration examples described above is recorded on a signal recording medium or transmitted via a transmission path. become.
[0182]
FIG. 22 illustrates an example in which an optical disc 704 is used as an example of a signal recording medium. In FIG. 22, a terminal 700 is supplied with a data sequence including the coded bit stream and information necessary for subsequent decoding such as a quantization scale. This data sequence is added with an error correction code by the ECC encoder 701 and sent to the modulation circuit 702. The modulation circuit 702 performs predetermined modulation processing, for example, processing such as 8-14 modulation, on the output of the ECC encoder 701. The output of the modulation circuit 702 is sent to a recording head 703, and is recorded on the optical disk 704 by the recording head 703.
[0183]
In the example of FIG. 22, an optical disk is taken as an example of a signal recording medium. However, a tape-shaped recording medium such as a magnetic tape, a magnetic disk medium such as a hard disk or a flexible disk, or a semiconductor storage medium such as an IC card or various memory elements is used. It is also possible to record a signal encoded by the device of the present invention on a signal recording medium such as a medium. Examples of the optical disk include a disk recorded by pits, a magneto-optical disk, a phase change optical disk, an organic dye optical disk, an optical disk recorded by ultraviolet laser light, and an optical disk having a multilayer recording film. Various discs can be used.
[0184]
【The invention's effect】
As is clear from the above description, in the present invention, the input moving image is divided into small blocks each including at least one pixel, the encoding difficulty of the small blocks is calculated by a predetermined method, and the encoding of the small blocks is performed. Based on the difficulty level and the filter control information used in the small block that the small block refers to in the inter-picture prediction, a filter characteristic for adaptively performing the low-pass filter processing on the small block is determined. In addition, the moving image subjected to the low-pass pre-filter processing can have a subjective impression of the encoded image quality and the encoding efficiency higher than before. That is, according to the present invention, in a moving picture coding apparatus having a configuration in which a pre-filter is provided in a stage preceding an inter-frame predictive coding circuit, the characteristics of a low-pass pre-filter can be partially changed in a frame, and a prediction image can be changed. It is possible to reduce the decrease in the coding efficiency due to the difference in the band of the image signal between them.
[0185]
In the present invention, the input moving image is divided into small blocks each including at least one pixel, the input image is classified into a filter characteristic image and a non-filter characteristic image, and the small block of the filter characteristic image is encoded by a predetermined method. Difficulty is calculated, and in the filter characteristic image, based on the coding difficulty of this small block and the filter control information used in the small block referred to in inter-picture prediction, Filter characteristic when adaptively performing low-pass filter processing, and in a non-filter characteristic image, a predetermined method is used to interpolate between the past reference image and the future reference image using a predetermined method. By calculating the filter characteristics when the low-pass filter processing is adaptively performed, the moving image subjected to the low-pass pre-filter processing is subjected to Than, better subjective impression of the coded image quality, and it is possible to have good coding efficiency. That is, according to the present invention, in a moving picture coding apparatus having a configuration in which a pre-filter is provided in a stage preceding an inter-frame predictive coding circuit, the characteristics of a low-pass pre-filter can be partially changed in a frame, and a prediction image can be changed. It is possible to reduce the decrease in the coding efficiency due to the difference in the band of the image signal between them.
[Brief description of the drawings]
FIG. 1 is a block circuit diagram illustrating a first configuration example of a moving picture encoding device according to the present invention.
FIG. 2 is a diagram showing a relationship between a frame and a small block.
FIG. 3 is a flowchart illustrating an example of a pre-filter control method.
FIG. 4 is a diagram showing a correspondence relationship between an average value d_ave of encoding difficulty levels and a representative value k_gop of prefilter coefficients.
FIG. 5 is a diagram showing an example of calculating a pre-filter coefficient in consideration of a pre-filter characteristic used in a small block referred to in a current small block inter-frame prediction.
FIG. 6 is a diagram illustrating a correspondence relationship between an encoding difficulty d_current and a low-pass pre-filter coefficient k_current.
FIG. 7 is a flowchart illustrating an example of a method for calculating a pre-filter coefficient k_current.
FIG. 8 is a diagram illustrating a calculation example of a low-pass pre-filter coefficient k_current.
FIG. 9 is a block circuit diagram illustrating a configuration example of a moving image encoding circuit in the moving image encoding device of the configuration example of the present invention.
FIG. 10 is a block circuit diagram showing a configuration example of an encoding difficulty measuring circuit in the video encoding device of the configuration example of the present invention.
FIG. 11 is a block circuit diagram showing a second configuration example of the video encoding device of the present invention.
FIG. 12 is a diagram showing a prediction structure of the MPEG system and a configuration of each picture type.
FIG. 13 is a block circuit diagram showing a third configuration example of the video encoding device of the present invention.
FIG. 14 is a diagram showing a display order and an encoding order in the MPEG system.
FIG. 15 is a diagram illustrating a relationship between frames and small blocks of a filter characteristic image and a non-filter characteristic image.
FIG. 16 is a flowchart illustrating an example of a pre-filter control method.
FIG. 17 is a diagram illustrating an example of calculating a pre-filter coefficient by interpolation using a pre-filter characteristic used in a small block referred to in the inter-frame prediction of a current small block in a non-filter characteristic image. is there.
FIG. 18 is a flowchart illustrating an example of a method for calculating a pre-filter coefficient k_current.
FIG. 19 is a diagram illustrating a calculation example of a low-pass pre-filter coefficient k_current of a non-filter characteristic image.
FIG. 20 is a diagram illustrating a configuration example of a prefilter.
FIG. 21 is a block circuit diagram illustrating a fourth configuration example of the video encoding device of the present invention.
FIG. 22 is a block circuit diagram showing a configuration for recording an encoded bit stream on an optical disc as an example of a signal recording medium.
FIG. 23 is a block circuit diagram illustrating a configuration example of a conventional video encoding device.
FIG. 24 is a diagram showing coefficients of a pre-low-pass filter.
FIG. 25 is a diagram illustrating a relationship between pre-filter coefficients and encoding difficulty.
[Explanation of symbols]
21, 91, 121, 191 frame memory group, 22, 92, 122, 192 prefilter, 23, 93, 123, 193 video coding circuit, 24, 96, 124, 196 motion vector detection circuit, 25, 95, 125, 195 pre-filter control information storage circuit, 26, 126 encoding difficulty measurement circuit, 27, 97, 127, 197 pre-filter control circuit, 301 low-pass filter, 302, 303, 304 arithmetic unit

Claims (30)

Dividing the input moving image into small blocks consisting of at least one pixel,
Calculate the encoding difficulty of the small block of the input moving image by a predetermined method,
The representative filter characteristic indicating a low-pass filter characteristic having a narrower pass band as the encoding difficulty of the small block of the input moving image is larger,
Based on the representative filter characteristics and filter characteristic control information of a low-pass filter process used before the encoding process when encoding the small block referred to in the inter-picture prediction by the small block of the input moving image. To determine filter characteristics when adaptively performing low-pass filter processing on the small blocks of the input moving image,
Along with performing an encoding process on the moving image signal that has been low-pass filtered with the determined filter characteristics,
The first band limit value, which is the band limit value of the low-pass filter in the filter characteristics used for the determined small block of the input moving image, and the small block to which the small block of the input moving image refers in inter-picture prediction are encoded. At the time of processing, a second band limit value which is a band limit value in a filter characteristic of a low-pass filter used before the encoding process is compared,
When the change amount of the first and second band limit values is larger than a predetermined threshold, the first band limit value is controlled so that the change amount is kept within the threshold value. Moving image encoding method. Performing predictive coding and / or orthogonal transform coding on the input moving image to generate coefficient data;
By calculating the coefficient data in a fixed quantization step to determine the generated code amount of the small block unit,
2. The moving picture coding method according to claim 1, wherein the generated code amount obtained in the small block unit is set as the coding difficulty. The prediction residual is obtained by detecting a motion vector between images of the input moving image in units of the small blocks,
2. The moving picture coding method according to claim 1, wherein the prediction residual obtained in the small block unit is used as the coding difficulty. The small block referred to in the inter-picture prediction by the small block of the input moving image is a small block located at a position referred to by a motion vector of the small block of the input moving image on a frame referred to in the inter-frame prediction. The moving picture encoding method according to claim 1, wherein: The small block referred to in the inter-picture prediction by the small block of the input moving image is a small block located at a position referred to by a motion vector of the small block of the input moving image on a frame referred to in the inter-frame prediction. The moving picture encoding method according to claim 1, wherein: The small block referred to in the inter-picture prediction by the small block of the input moving image is a small block located at the same spatial position as the small block of the input moving image on a frame referred to in the inter-frame prediction. The moving picture coding method according to claim 1, wherein The small block referred to in the inter-picture prediction by the small block of the input moving image is a small block located at the same spatial position as the small block of the input moving image on a frame referred to in the inter-frame prediction. The moving picture coding method according to claim 1, wherein Dividing means for dividing the input moving image into small blocks each including at least one pixel;
Calculating means for calculating the encoding difficulty of the small block of the input moving image by a predetermined method,
As the encoding difficulty of the small block of the input moving image increases, a representative filter characteristic indicating a low-pass filter characteristic having a narrower pass band is obtained, and the representative filter characteristic and the small block of the input moving image are referred to in inter-picture prediction. When the small block is encoded, the low-pass filter processing is adaptively performed on the small block of the input moving image based on the filter characteristic control information of the low-pass filter processing used before the encoding processing. Filter characteristics determining means for determining the filter characteristics at the time of,
Encoding means for performing an encoding process on the moving image signal subjected to the low-pass filter processing with the determined filter characteristics,
The filter characteristic determining means includes: a first band limit value that is a band limit value of a low-pass filter in a filter characteristic used for the determined small block of the input moving image; The first and second band limit values are compared with a second band limit value which is a band limit value of a low-pass filter used before the encoding process when encoding the small block to be referred to. Comparing means for determining the amount of change of the first band limit value and the first band limit value such that the amount of change of the first and second band limit values is smaller than the threshold value if the change amount is larger than a predetermined threshold value. A moving image encoding apparatus comprising: a control unit that controls the moving image. The calculating means performs predictive coding and / or orthogonal transform coding on the input moving image to generate coefficient data, and quantizes the coefficient data in a fixed quantization step 9. A moving image according to claim 8, further comprising: a generated code amount generating means for obtaining the generated code amount in small block units, wherein the generated code amount obtained in small block units is used as the encoding difficulty. Encoding device. The calculating means includes a prediction residual generating means for detecting a motion vector between images of the input moving image in small block units to obtain a prediction residual, and calculating the prediction residual obtained in the small block unit by the code. 9. The moving picture coding apparatus according to claim 8, wherein the moving picture coding difficulty is set. The small block referred to in the inter-picture prediction by the small block of the input moving image is a small block located at a position referred to by a motion vector of the small block of the input moving image on a frame referred to in the inter-frame prediction. 9. The moving picture coding apparatus according to claim 8, wherein: The small block referred to in the inter-picture prediction by the small block of the input moving image is a small block located at a position referred to by a motion vector of the small block of the input moving image on a frame referred to in the inter-frame prediction. 9. The moving picture coding apparatus according to claim 8, wherein: The small block referred to in the inter-picture prediction by the small block of the input moving image is a small block located at the same spatial position as the small block of the input moving image on a frame referred to in the inter-frame prediction. The moving picture coding apparatus according to claim 8, wherein The small block referred to in the inter-picture prediction by the small block of the input moving image is a small block located at the same spatial position as the small block of the input moving image on a frame referred to in the inter-frame prediction. The moving picture coding apparatus according to claim 8, wherein For the input moving image, adaptively low-pass filter processing is performed on a reference image that is referred to when determining filter characteristics when adaptively performing low-pass filter processing on an intra-frame encoded image, and on an inter-frame encoded image. Classification into non-reference images that are not referenced when determining the filter characteristics at the time of
Dividing the input moving image into small blocks consisting of at least one pixel,
When the input moving image is a reference image of the filter characteristics, a low-pass filter characteristic having a narrower pass band as the encoding difficulty of the small block in the input moving image calculated by a predetermined method is larger is selected and selected. The low-pass filter characteristic is modified based on the filter control information used in the small block referred to in the inter-picture prediction by the small block of the input moving image, and the low-pass filtering is adaptively performed on the small block of the input moving image Determine the filter characteristics when applying
When the input moving image is a non-reference image having filter characteristics, filter control information of a small block in a past reference image referred to by a small block of the input moving image in inter-picture prediction, and filter control of a small block in a future reference image Based on the information, a filter characteristic when adaptively performing a low-pass filter process on the small block of the input moving image is determined,
Along with performing an encoding process on the moving image signal that has been low-pass filtered with the determined filter characteristics,
When the input moving image is a reference image having a filter characteristic, the coding difficulty of all the small blocks in the input moving image is smoothed in the input moving image. By comparing a first band limit value that is a value and a second band limit value that is a band limit value of a low-pass filter used in a small block referred to in inter-picture prediction with a small block of the input moving image. The amount of change of the first and second band limit values is obtained, and when the amount of change is greater than a predetermined threshold, the first band limit value is set so as to keep the amount of change within the threshold. Control and
When the input moving image is a non-reference image having a filter characteristic, a small block of the input moving image is a band limit value of a low-pass filter used in a small block in a past reference image referred to in inter-picture prediction. From the band limit value of No. 3 and the fourth band limit value of the low-pass filter used for the small block in the future reference image, the first band limit value is calculated by interpolation using a predetermined method. Comparing the first band limit value with a fifth band limit value of a low-pass filter used in a small block at the same spatial position in an image temporally close to the input moving image. To determine the amount of change in the first and fifth band limit values,
When the amount of change is larger than a predetermined threshold, the first band limit value is controlled so that the amount of change is kept within the threshold. The reference image is subjected to predictive coding and / or orthogonal transform coding to generate coefficient data, and the coefficient data is quantized in a fixed quantization step to obtain the generated code amount in the small block unit. 16. The moving picture coding method according to claim 15, wherein the generated code amount obtained in the small block unit is set as the coding difficulty. 16. The method according to claim 15, further comprising: detecting a motion vector between the images of the reference image in the small block unit to obtain a prediction residual; and setting the prediction residual obtained in the small block unit as the encoding difficulty. The moving picture coding method according to the above. The small block referred to in the inter-picture prediction by the small block of the input moving image is a small block at a position referred to by the motion vector of the small block of the input moving image on a frame referred to in the inter-frame prediction. The moving picture encoding method according to claim 15, wherein: The small block referred to in the inter-picture prediction by the small block of the input moving image is a small block at a position referred to by the motion vector of the small block of the input moving image on a frame referred to in the inter-frame prediction. The moving picture encoding method according to claim 15, wherein: The small block referred to in the inter-picture prediction by the small block of the input moving image is a small block located at the same spatial position as the small block of the input moving image on a frame referred to in the inter-frame prediction. The moving picture coding method according to claim 15, wherein The small block referred to in the inter-picture prediction by the small block of the input moving image is a small block located at the same spatial position as the small block of the input moving image on a frame referred to in the inter-frame prediction. The moving picture coding method according to claim 15, wherein Dividing means for dividing the input moving image into small blocks each including at least one pixel;
Classification means for classifying the input video into a reference image having a filter characteristic of an intra-frame coded image and a non-reference image having a filter characteristic of an inter-frame image,
When the input moving image is a reference image having filter characteristics, a calculating unit that calculates a coding difficulty of a small block of the input moving image by a predetermined method,
Encoding difficulty of the small blocks of the input video, smoothing means for smoothing the encoding difficulty of all small blocks in the input video,
When the input moving image is a reference image of the filter characteristics, a low-pass filter characteristic with a narrower pass band is selected as the encoding difficulty of the small block in the input moving image calculated by a predetermined method is larger, and the selected is selected. The low-pass filter characteristic is modified based on the filter control information used in the small block referred to in the inter-picture prediction by the small block of the input moving image, and the low-pass filter processing is adaptively performed on the small block of the input moving image. Is determined, and when the input moving image is a non-reference image having the filter characteristics, the filter control information of the small block in the past reference image to which the small block of the input moving image is referred to in the inter-picture prediction And the low-pass adaptively to the small block of the input moving image based on the filter control information of the small block in the future reference image. A filter characteristic determining means for determining the filter characteristics when subjected to filter processing,
Encoding processing means for performing an encoding process on the moving image signal subjected to the low-pass filter processing with the determined filter characteristics,
The filter characteristic determination means, when the input moving image is a non-reference image of the filter characteristics, a small block of the input moving image refers to a low-pass filter used in a small block in a past reference image referred to in inter-picture prediction. From the third band limit value that is the band limit value and the fourth band limit value that is the band limit value of the low-pass filter used for the small block in the future reference image, the first band limit value is obtained by interpolation. The interpolation interpolating means to calculate, and when the input moving image is a reference image of a filter characteristic, the encoding difficulty of all the small blocks in the input moving image is smoothed and calculated in the input moving image. A first band limit value that is a band limit value of a low-pass filter of a small block, and a low-pass filter used in a small block that the small block of the input moving image refers to in inter-picture prediction By comparing the first and second band limit values with a second band limit value that is a band limit value, and when the input moving image is a non-reference image with filter characteristics, The first band limit value obtained by the interpolation unit and the fifth band limit value of the low-pass filter used in the small block at the same spatial position in the image temporally close to the input moving image. Comparing means for comparing the first and fifth band limit values with the band limit value, and when the change amount obtained by the comparing means is larger than a predetermined threshold value, A moving image coding apparatus comprising: a control unit that controls the first band limit value so as to keep the value within the threshold value. The calculating means includes coefficient data generating means for performing predictive coding and / or orthogonal transform coding on the input moving image to generate coefficient data, and quantizing the coefficient data in a fixed quantization step. 23. A moving image according to claim 22, further comprising: a generated code amount generating means for calculating the generated code amount in small block units, wherein the generated code amount obtained in small block units is used as the encoding difficulty. Encoding device. The calculation means includes a prediction residual generation means for detecting a motion vector between images of the input moving image in units of the small blocks to obtain a prediction residual, and calculating the prediction residual obtained in units of the small blocks by the code 23. The moving picture coding apparatus according to claim 22, wherein the moving picture coding difficulty is set. The small block referred to in the inter-picture prediction by the small block of the input moving image is a small block located at a position referred to by a motion vector of the small block of the input moving image on a frame referred to in the inter-frame prediction. 23. The moving picture coding apparatus according to claim 22, wherein: The small block referred to in the inter-picture prediction by the small block of the input moving image is a small block located at a position referred to by a motion vector of the small block of the input moving image on a frame referred to in the inter-frame prediction. 23. The moving picture coding apparatus according to claim 22, wherein: The small block referred to in the inter-picture prediction by the small block of the input moving image is a small block located at the same spatial position as the small block of the input moving image on a frame referred to in the inter-frame prediction. 23. The moving picture coding apparatus according to claim 22, wherein: The small block referred to in the inter-picture prediction by the small block of the input moving image is a small block located at the same spatial position as the small block of the input moving image on a frame referred to in the inter-frame prediction. 23. The moving picture coding apparatus according to claim 22, wherein: For an input moving image encoded by the MPEG method, an I-picture and a B-picture are adaptively applied to a low-pass filter to determine a filter characteristic, or a P-picture is adaptively applied to a low-pass filter. Classify into non-reference images that are not referenced when determining filter characteristics when performing filter processing,
Dividing the input moving image into small blocks consisting of at least one pixel,
When the input moving image is a reference image of the filter characteristics, a low-pass filter characteristic having a narrower pass band as the encoding difficulty of the small block in the input moving image calculated by a predetermined method is larger is selected and selected. The low-pass filter characteristic is modified based on the filter control information used in the small block referred to in the inter-picture prediction by the small block of the input moving image, and the low-pass filtering is adaptively performed on the small block of the input moving image Determine the filter characteristics when applying
When the input moving image is a non-reference image having filter characteristics, filter control information of a small block in a past reference image referred to by a small block of the input moving image in inter-picture prediction, and filter control of a small block in a future reference image Based on the information, a filter characteristic when adaptively performing a low-pass filter process on the small block of the input moving image is determined,
Along with performing an encoding process on the moving image signal that has been low-pass filtered with the determined filter characteristics,
When the input moving image is a reference image having a filter characteristic, the coding difficulty of all the small blocks in the input moving image is smoothed in the input moving image. By comparing a first band limit value that is a value and a second band limit value that is a band limit value of a low-pass filter used in a small block referred to in inter-picture prediction with a small block of the input moving image. The amount of change of the first and second band limit values is obtained, and when the amount of change is greater than a predetermined threshold, the first band limit value is set so as to keep the amount of change within the threshold. Control and
When the input moving image is a non-reference image having a filter characteristic, a small block of the input moving image is a band limit value of a low-pass filter used in a small block in a past reference image referred to in inter-picture prediction. From the band limit value of No. 3 and the fourth band limit value of the low-pass filter used for the small block in the future reference image, the first band limit value is calculated by interpolation using a predetermined method. Comparing the first band limit value with a fifth band limit value of a low-pass filter used in a small block at the same spatial position in an image temporally close to the input moving image. To determine the amount of change in the first and fifth band limit values,
When the amount of change is larger than a predetermined threshold, the first band limit value is controlled so that the amount of change is kept within the threshold. Dividing means for dividing the input moving image encoded by the MPEG method into small blocks each including at least one pixel;
A classification unit that classifies the I picture and the B picture into a reference image having a filter characteristic and the P picture into a non-reference image having a filter characteristic in the input moving image;
When the input moving image is a reference image having filter characteristics, a calculating unit that calculates a coding difficulty of a small block of the input moving image by a predetermined method,
Encoding difficulty of the small blocks of the input video, smoothing means for smoothing the encoding difficulty of all small blocks in the input video,
When the input moving image is a reference image of the filter characteristics, a low-pass filter characteristic with a narrower pass band is selected as the encoding difficulty of the small block in the input moving image calculated by a predetermined method is larger, and the selected is selected. The low-pass filter characteristic is modified based on the filter control information used in the small block referred to in the inter-picture prediction by the small block of the input moving image, and the low-pass filter processing is adaptively performed on the small block of the input moving image. Is determined, and when the input moving image is a non-reference image having the filter characteristics, the filter control information of the small block in the past reference image to which the small block of the input moving image is referred to in the inter-picture prediction And the low-pass adaptively to the small block of the input moving image based on the filter control information of the small block in the future reference image. A filter characteristic determining means for determining the filter characteristics when subjected to filter processing,
Encoding processing means for performing an encoding process on the moving image signal subjected to the low-pass filter processing with the determined filter characteristics,
The filter characteristic determination means, when the input moving image is a non-reference image of the filter characteristics, a small block of the input moving image refers to a low-pass filter used in a small block in a past reference image referred to in inter-picture prediction. From the third band limit value that is the band limit value and the fourth band limit value that is the band limit value of the low-pass filter used for the small block in the future reference image, the first band limit value is obtained by interpolation. The interpolation interpolating means to calculate, and when the input moving image is a reference image of a filter characteristic, the encoding difficulty of all the small blocks in the input moving image is smoothed and calculated in the input moving image. A first band limit value that is a band limit value of a low-pass filter of a small block, and a low-pass filter used in a small block that the small block of the input moving image refers to in inter-picture prediction By comparing the first and second band limit values with a second band limit value that is a band limit value, and when the input moving image is a non-reference image with filter characteristics, The first band limit value obtained by the interpolation unit and the fifth band limit value of the low-pass filter used in the small block at the same spatial position in the image temporally close to the input moving image. Comparing means for comparing the first and fifth band limit values with the band limit value, and when the change amount obtained by the comparing means is larger than a predetermined threshold value, A moving image coding apparatus comprising: a control unit that controls the first band limit value so as to keep the value within the threshold value.

Images