JP2006186979A - Dynamic image conversion apparatus - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a dynamic image conversion apparatus which is capable of performing data conversion between different encoding methods, in a short time, such as from MPEG-4/AVC standard to MPEG-2 standard or from MPEG-4/AVC standard to MPEG-4 standard. <P>SOLUTION: The dynamic image conversion apparatus converts first dynamic image data encoded in accordance with a first motion compensated prediction method into second dynamic image data that have a same format as data encoded in accordance with a second motion compensated prediction method. The dynamic image conversion apparatus determines whether a relationship between a block in the second motion compensated prediction method and an image used as a reference image with respect to the block conforms with a condition. For the block of the relationship determined to conform with the condition, the dynamic image conversion apparatus performs encoding using a motion vector or vectors of the first dynamic image data corresponding to the block. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention relates to a moving image conversion apparatus that converts a code sequence of a coded moving image, and more particularly to improvement of conversion efficiency in conversion of a moving image encoded by a different motion compensation prediction method into a code sequence.

In recent years, digitalization of moving images has progressed, and moving images are transmitted as digital data over a network or the like and recorded on a medium.
In order to meet the demands for the color, definition, and smoothness of motion of these moving images, the amount of data has increased, and although the communication line speed has increased and the storage medium has increased in capacity, A compression technique was sought.

Therefore, for example, the MPEG-2 standard with the goal of realizing high image quality such as television broadcasting from the current television to the high-definition, and the MPEG-4 standard with the goal of realizing a high compression ratio for use on mobile and the Internet. Coding compression methods such as were developed.
These have applications according to their characteristics, but it is desirable for users to be able to convert moving images to each other. This is because, for example, a television program recorded according to the MPEG-2 standard can be converted into MPEG-4 standard data and viewed on a portable playback device.

Usually, in these data conversions, data before conversion, for example, data encoded according to the MPEG-2 standard is once completely decoded, and the decoded moving image is converted into a data format after conversion, for example, according to the MPEG-4 standard. This is done by re-encoding.
However, although this conversion method has the advantage that it can be realized by connecting a conventional encoding device and decoding device, it basically takes time for the playback time of a moving image, so a HDD (Hard Disk For users who are accustomed to high-speed dubbing from a drive (DVD) to a DVD (Digital Versatile Disk) or the like, the convenience is greatly deteriorated.

Therefore, the data conversion time has been shortened by reusing part of the encoded data before conversion.
This conversion method is based on the premise that the encoding process before and after the conversion is basically the same. For example, in data conversion from MPEG-2 to its higher standard MPEG-4, motion vectors are reused. By reducing the processing time required for the motion vector search process, the conversion time is greatly shortened.

However, in the opposite case, that is, data conversion from MPEG-4 to MPEG-2, motion vectors may not be reused.
This is because MPEG-2 does not support all specifications related to block sizes, motion vectors, and the like that are supported by the higher standard MPEG-4. Specifically, the block size supported by MPEG-2 is only 16 pixels × 16 pixels, but MPEG-4 additionally supports blocks of 8 pixels × 8 pixels. Furthermore, in MPEG-2, motion vectors are allowed only up to ½ pixel accuracy, but in MPEG-4, up to ¼ pixel accuracy is allowed.

  A technique for performing data conversion by solving these differences has been developed (Patent Document 1). Briefly describing the solution, one motion vector of a corresponding 16 pixel × 16 pixel block is calculated from four motion vectors obtained from four 8 pixel × 8 pixel blocks, and with a 1/4 pixel accuracy. Data conversion is realized by performing a re-search in the vicinity of the motion vector and obtaining and using a motion vector with 1/2 pixel accuracy.

Recently, in addition to the MPEG-2 standard and the MPEG-4 standard, the MPEG-4 / AVC standard (Non-Patent Document 1) whose encoding processing is basically the same as these standards has been newly standardized.
This MPEG-4 / AVC standard can also be applied to a mobile or the Internet targeted by MPEG-4 from a television broadcast or the like targeted by MPEG-2. QCIF (Quarter CIF: 180 pixels × 144 lines) to HD (High Definition: 1920 pixels × 1080 lines) is a scalable standard that can handle images up to a size, and in addition, a standard that realizes an extremely high compression rate compared to other standards.

Therefore, if the MPEG-4 / AVC standard becomes widespread, it is predicted that the demand for moving image conversion devices between the MPEG-2 standard and the MPEG-4 standard, which have high affinity with conventional products such as DVDs, will increase. Is done.
JP-A-11-275592 ITU-TH. H.264 standard

However, since the MPEG-4 / AVC standard has a different part from MPEG-4, data conversion from MPEG-4 / AVC to MPEG-2 cannot be performed even with the above technology. The time for the conversion process cannot be shortened.
Accordingly, the present invention provides a moving image capable of performing data conversion between different encoding methods such as MPEG-4 / AVC standard to MPEG-2 standard and MPEG-4 / AVC standard to MPEG-4 standard in a short time. An object is to provide a conversion device.

  In order to achieve the above object, a moving image conversion apparatus according to the present invention is the same as data encoded according to a second motion compensated prediction method, in which first moving image data encoded according to a first motion compensated prediction method is the same. A moving image conversion apparatus for converting into a second moving image data in a format, wherein one or more motion vectors of a block included in the first moving image data are motion vectors in the second motion compensation prediction method of the block. Selection means for selecting a motion vector that satisfies the condition to be satisfied, and encoding means for performing encoding according to the second motion compensation prediction method, wherein each motion vector selected by the selection means is used to block And an encoding means for performing the encoding.

  When the motion vector of the block encoded by the encoding method before the conversion is selected as the condition of the encoding method after the conversion is selected, the moving image conversion device having the above configuration is the motion vector before the conversion. Since the encoding is performed by the encoding method after conversion using, there is no need to obtain a motion vector by a conventional motion search, and the data conversion processing time can be shortened.

In addition, in the encoding using the selected motion vectors, the encoding means performs the encoding using the selected motion vectors as they are instead of using the motion vector specified by the motion search. It is good as well.
As a result, since the motion vector before conversion can be used as it is without performing any motion search, the time required for the conversion process can be greatly reduced.

  In the encoding using each selected motion vector, the encoding means is a predetermined search range based on a position on a reference image indicated by each motion vector, and the motion vector indicates Instead of performing a motion search for a search range including a reference block and using the motion vector, a block may be encoded using a motion vector obtained as a result of the motion search.

As a result, only the predetermined search range based on the motion vector before conversion is searched to obtain the motion vector, so that the search range is narrowed compared to the conventional case where the search is performed from a wide range without relying on the motion vector. As a result, the amount of calculation required for the search can be reduced, and the time required for the conversion process can be greatly shortened.
Further, the condition to be satisfied by the motion vector selected by the selection means is to limit the encoding type of the reference image of the block, and the selection means may select a motion vector that satisfies the condition.

The condition to be satisfied by the motion vector selected by the selection means is to limit the encoding type of the reference image of the block, and the selection means refers to the motion vector of the block included in the first moving image data. When the image does not satisfy the condition, the motion vector may be selected by replacing the reference image with a reference image that satisfies the condition.
As a result, the motion vector before the conversion is used as it is, the reference block in the reference image that is allowed to be referred to in the encoded encoding method is used as the prediction signal, and the reference is allowed in the encoded encoding method. Since the reference block in the reference image is used as the prediction signal, the calculation for the motion vector search becomes unnecessary, and the processing time can be shortened.

  In the encoding using each selected motion vector, the encoding means is a predetermined search range based on a position on a reference image indicated by each motion vector, and the motion vector indicates Instead of performing a motion search for a search range including a reference block and using the motion vector, a block may be encoded using a motion vector obtained as a result of the motion search.

Accordingly, a motion vector is obtained by searching a predetermined search range based on the motion vector in the first moving image before conversion on another reference image that is allowed to be referenced. The search range can be narrowed down compared to the case, and the processing time can be shortened.
The condition to be satisfied by the motion vector selected by the selection means is that when the motion vector of the block is one, the reference image of the motion vector is an image displayed before the image including the block. In other words, the selection means may select a motion vector when one of the motion vectors of the block included in the first moving image data satisfies the condition.

  The condition that the motion vector selected by the selection means should satisfy is that when the motion vector of a block is two, the reference image of one of the motion vectors is displayed before the image including the block. The reference image of the other motion vector is an image that is displayed after the image including the block, and the selecting means is a motion of the block included in the first moving image data. If two of the vectors satisfy the condition, the two motion vectors may be selected.

  The condition that the motion vector selected by the selection means should satisfy is that the encoding type of the image displayed between the image including the block and the reference image of the block is a predetermined type in the image display order. In other words, the selection means may select the motion vector when the reference image of the motion vector of the block included in the first moving image data satisfies the condition.

  The condition to be satisfied by the motion vector selected by the selection means is that the reference image of the block is the image of the past P picture or I picture that is closest in display time order from the display of the image including the block, or the most It is an image of a P picture or an I picture in the near future, or an image including the block, and the selection unit is configured such that the reference image of the motion vector of the block included in the first moving image data satisfies the condition Alternatively, the motion vector may be selected.

  Further, when the number of motion vectors of the block included in the first moving image data is two, and the selecting means selects only one motion vector that satisfies the condition, the encoding means Encoding a block using a motion vector and changing the information indicating the motion prediction type of the block from a motion prediction type using two motion vectors to a motion prediction type using one motion vector It is good.

Further, the condition is that all the reference blocks of the block are present in the reference image including the reference block, and the selecting means includes the reference block of the motion vector of the block included in the first moving image data. If the condition is satisfied, the motion vector may be selected.
In addition, if the reference block of the motion vector after the correction exists in the reference image by correcting the motion vector among the motion vectors that have not been selected, A motion vector may be selected.

Thereby, since the motion vector suitable for the encoding system after conversion can be selected among the motion vectors before conversion, the selected motion vector can be reused and encoded.
Further, the first motion compensation prediction method may be based on the MPEG4-AVC standard.
Thereby, the data encoded by the MPEG-4 / AVC standard can be converted into the same format as the data encoded by another standard.

Further, the second motion compensation prediction method may be based on the MPEG2 standard or the MPEG4 standard.
As a result, data encoded according to a standard such as the MPEG-4 / AVC standard can be converted into the same format as the data encoded according to the MPEG-2 standard.
Also, a moving image conversion method for converting the first moving image data encoded according to the first motion compensation prediction method into the second moving image data having the same format as the data encoded according to the second motion compensation prediction method is provided. Selecting a motion vector satisfying a condition to be a motion vector in the second motion compensation prediction method of the block from one or more motion vectors of the block included in the first moving image data; An encoding step for performing encoding according to the motion compensated prediction method, wherein the block is encoded using each motion vector selected in the selection step.

  In addition, the second moving image having the same format as the data encoded according to the second motion compensation prediction method is converted from the first moving image data encoded according to the first motion compensation prediction method to an apparatus capable of executing a computer program. The computer program for performing the moving image conversion process for converting into data becomes a motion vector in the second motion compensation prediction method of the block from one or more motion vectors of the block included in the first moving image data. A selection step for selecting a motion vector satisfying a power condition, and an encoding step for performing encoding according to the second motion compensation prediction method, wherein each motion vector selected in the selection step is used to code a block And an encoding step for performing the conversion.

  A computer-readable recording medium having a computer program recorded thereon, wherein the computer program stores, on a device capable of executing the computer program, first moving image data encoded according to a first motion compensation prediction method. A computer program for performing a moving image conversion process for converting into second moving image data in the same format as data encoded according to the motion compensation prediction method of No. 2, wherein one or more blocks included in the first moving image data A selection step for selecting a motion vector satisfying a condition to be a motion vector in the second motion compensated prediction method for the block, and a code for performing encoding according to the second motion compensated prediction method Using the motion vectors selected in the selection step. Characterized in that it comprises an encoding step of performing-coding.

  In addition, a moving image conversion apparatus that converts first moving image data encoded according to a first motion compensation prediction method into second moving image data having the same format as data encoded according to a second motion compensation prediction method. An integrated circuit that selects, from one or more motion vectors of a block included in the first moving image data, a motion vector that satisfies a condition to be a motion vector in the second motion compensation prediction method of the block; Encoding means for performing encoding according to a second motion compensation prediction method, and encoding means for encoding a block using each motion vector selected by the selecting means. And

<Embodiment 1>
<Overview>
A moving image conversion apparatus according to the present invention is an apparatus that converts data encoded according to the MPEG-4 / AVC standard into a data format encoded according to the MPEG-2 standard, and a motion vector is converted during the conversion. It is characterized by reuse. Since these encoding methods share the same basic encoding in that they are motion compensated prediction methods, motion vectors that are common components of encoded data are reused.

Since the MPEG encoding process has a feature that the amount of motion vector search processing becomes very large, reducing the motion vector search process significantly reduces the encoding processing time and greatly improves convenience. become.
First, the difference in motion vectors between the MPEG-4 / AVC standard and the MPEG-2 standard will be described with reference to FIGS.

<Differences between MPEG-4 / AVC standard and MPEG2 standard>
MPEG-4 / AVC and MPEG-2 video streams will be briefly described with reference to the drawings. These video streams have basically the same shape.
<Video stream>
FIG. 23 is a diagram illustrating a data structure of a video stream.

The video stream is compressed based on the spatial frequency component of the image, and several pixels on the screen are used as one compression unit. The compression unit is called a macro block, and the screen is composed of a plurality of macro blocks. FIG. 24 is a diagram in which one screen is composed of macroblocks (0, 0) to (i, j).
This macroblock is composed of, for example, 16 pixels vertically × 16 pixels horizontally, and decoding processing is performed in units of one at the time of decoding.

The first level in FIG. 23 shows the configuration of the video stream.
The video stream is composed of a plurality of GOPs (Group of Picture) (see the first figure), and the GOP has three types of picture data (I, P in the figure): I picture, B picture, and P picture. , B) including one or more, and an I picture always exists at the head (see the second figure).
Further, each picture data is composed of a picture header and a plurality of slices (see the third stage figure), and the picture header includes a picture type. This picture type indicates whether the picture data is an I picture, a B picture, or a P picture.

However, in MPEG-4 / AVC, the picture type itself is not included, but the slice type included in the picture is included, and the picture type is obtained from these slice types.
A slice has a slice header and a plurality of macroblocks (see the fourth figure), and the slice header includes a slice type. This slice type indicates whether the slice data is an I slice, a B slice, or a P slice.

However, this slice type exists only in MPEG-4 / AVC data. That is, in MPEG-4 / AVC, there are multiple types of slices in a picture, but in MPEG-2, there is only one type of slice.
Next, the macro block 2300 includes a macro block type 2310, motion information 2320, and an inter-screen prediction error 2330.

The macro block type represents information such as whether or not this macro block requires inter-screen prediction and whether it is a one-way reference or a two-way reference, and the type of motion information differs depending on this type. It will be.
The motion information includes a reference picture 2321 that identifies a picture to be referred to and a motion vector 2322.

In the figure, an example of motion information with reference to two directions is shown. “Ref.idx0” and “ref.idx1” are numbers indicating reference pictures, and “mv0” and “mv1” represent motion vectors in the respective reference pictures.
However, this example of motion information is in the MPEG-4 / AVC standard, and in the MPEG-2 standard, there is no “ref.idx0” that is a number indicating a reference picture. This is because in MPEG-2, since a picture that can be referred to is determined, it is not necessary to have it as data.

In the present invention, an MPEG-2 motion vector is created by reusing an MPEG-4 / AVC motion vector, but the macroblock type and reference picture are changed depending on the available motion vector.
The inter-screen prediction error 2330 represents the difference between the reference image and the target image.
<I / P / B picture>
Here, the I picture is a screen obtained by encoding using only information in the own screen, the P picture is a screen obtained by one-way predictive encoding with one screen, and the B picture is A screen obtained by bi-directional predictive encoding with two screens.

  Further, all macroblocks included in the I picture are (1) macroblocks obtained only from information in the own screen, and macroblocks included in the P picture are (1) macros obtained only from information in the own screen. The block is composed of a block and (2) a macroblock obtained by one-way predictive coding with one screen, and the macroblock included in the B picture is (1) a macroblock obtained only with information in the own screen, and (2) It consists of a macroblock obtained by one-way predictive coding with one screen and (3) a screen obtained by two-way predictive coding with two screens. These types are represented by macroblock types (see FIG. 23).

Since there is a difference in the macro block types supported by MPEG-4 / AVC and MPEG-2, it can be said that the object of the present invention is to eliminate this difference. These differences will be described in <Motion vector difference> below.
In MPEG-4 / AVC, each slice of I / P / B is also composed of macroblocks of the same type as the picture described above, the I picture is composed of only I slices, and the P picture is composed of I slices and P slices. , B picture is composed of I slice, P slice and B slice. Accordingly, the MPEG-4 / AVC picture and the MPEG-2 picture each include the same type of macroblock, and therefore, in the present embodiment, description will be made using pictures.

In actual MPEG-4 / AVC, since processing is performed in units of slices, the present invention may be implemented in units of slices.
<Difference in motion vector>
The difference in motion vector between MPEG-4 / AVC and MPEG-2 is the difference in the degree of freedom of reference image selection. In MPEG-4 / AVC, the range of selection of reference images is expanded in order to increase the compression rate.

  FIG. 25 shows an example of a motion vector of MPEG-2, FIG. 25 (a) shows an example of a P picture, and FIG. 25 (b) shows an example of a B picture. FIG. 26 shows an example of a motion vector of MPEG-4 / AVC, and FIGS. 26A and 26B are examples of a B picture. In these figures, “I / P / B” or the like represents an I picture, a P picture, a B picture, or the like, and an arrow indicates a target block from the reference block (in FIG. 5 or the like). The same is true.)

First, the I picture can be converted between these standards because it is obtained by encoding independently only within the own screen without performing motion prediction in the time direction.
However, all of the macroblocks constituting the MPEG-2 I picture can be encoded only by the own macroblock (non-reference), but the macroblock constituting the MPEG-4 / AVC I picture is the non-macroblock. In addition to the reference macroblock, there is a macroblock that encodes the target block with reference to other macroblocks (upper left, upper, upper right, and left) in the own screen (intra prediction).

At the time of data conversion, the intra prediction type macroblock is converted into a non-reference type macroblock. In this case, since only the target macroblock needs to be encoded, the conversion can be performed with a small processing time.
Next, regarding the P picture and the B picture, there are three major differences between the standards.
One is that the types of pictures that can be used as reference images are different, two is the temporal distance between pictures that can be used as reference images, and three is the temporal direction.

In terms of picture types, MPEG-2 must be an I picture or a P picture (see FIG. 25A), but in MPEG-4 / AVC, a B picture can also be a reference picture. (See FIG. 26 (a)).
In addition, the temporal distance between pictures that can be used as reference images is limited to the temporally nearest picture in MPEG-2. For example, an MPEG-2 P picture needs to use “the closest past I picture or P picture” as a reference image (see FIG. 25A).

However, in MPEG-4 / AVC, in the case of a P picture, the latest 16 pictures can be used as reference images. For example, a B picture is a picture of 16 pictures in the past and future. A reference image can be used (see FIG. 26A).
Furthermore, as a restriction in the temporal direction, it is necessary to select one reference image forward and backward in the MPEG-2 B picture. For example, of the two motion vectors for a block of B picture, one of them is “the closest P picture or I picture that is closest in time order” as a reference image, and the other one is “the future P picture that is closest in time order” It is necessary to use “picture or I picture” as a reference image (see FIG. 25B).

However, in MPEG-4 / AVC, two images from the front direction or two images from the rear direction can be selected as reference images (see FIG. 26B).
These differences are summarized based on the macroblock type as follows.
In this table, “forward” means a forward reference that refers to a temporally forward picture, and “backward” means a backward reference that refers to a temporally backward picture. .

本発明は、これらの相違をふまえ、動きベクトルを再利用し、データの変換を行っている。

Based on these differences, the present invention reuses motion vectors and performs data conversion.

Hereinafter, a moving image conversion apparatus according to an embodiment of the present invention will be described with reference to the drawings.
<Function>
FIG. 1 is a block diagram illustrating a configuration of a moving image conversion apparatus.
In FIG. 1, a moving image conversion apparatus 100 is an apparatus that converts an MPEG-4 / AVC format moving image input from the outside into an MPEG-2 format moving image, and outputs the moving image in an MPEG-2 format. 20 and a motion vector configuration unit 30.

The decoding unit 10 receives MPEG-4 / AVC format moving image data as input and decodes it. Also, the decoding unit 20 receives the moving image decoded by the decoding unit 10 and encodes it into MPEG-2 format moving image data.
Here, the decoding unit 10 is basically the same as the conventional decoding device in each functional unit. However, there is a data flow for performing processing unique to the present invention (thick arrows), which will be described in the description of the motion vector configuration unit 30 described later.

Compared to the conventional encoding unit, the encoding unit 20 is characterized in that there is no functional unit that performs motion vector search. This is because the present invention is characterized by reusing motion vectors, so that it is not necessary to search for motion vectors as in the prior art.
Instead, a motion vector configuration unit 30 is newly provided.
The motion vector configuration unit 30 determines whether or not a motion vector included in a moving image in the MPEG-4 / AVC format can be reused, and the motion compensation prediction unit 28 determines whether to use data for reuse or the like according to the determination result. Send instructions.

Hereinafter, first, the configurations of the decoding unit 10 and the encoding unit 20 will be briefly described, and then the motion vector configuration unit 30 will be described.
<Decoding unit 10>
The decoding unit 10 includes a variable length decoding unit 11, an inverse quantization unit 12, an inverse orthogonal transform unit 13, an addition unit 14, a frame memory 15, and a motion compensation unit 16.

The variable length decoding unit 11 performs variable length decoding when a moving image code string encoded in the MPEG-4 / AVC format is input from the outside, and performs picture type, macroblock type, motion information, quantization coefficient Isolate etc.
The inverse quantization unit 12 and the inverse orthogonal transform unit 13 perform inverse quantization and inverse orthogonal transform on the separated quantization coefficient for each block to obtain pixel data for each block.

The adder 14 outputs the pixel data for each block as it is or adds the prediction signal from the motion compensator 16 depending on the type of encoding indicated by the macroblock type.
The frame memory 15 accumulates pixel data corresponding to a screen to be used as a reference image among outputs from the adding unit 14.
The motion compensation unit 16 obtains a prediction signal of the block from the motion information obtained from the variable length decoding unit 11 and the reference image stored in the frame memory 15 and adds the prediction signal to the motion compensation predicted block. 14 for output.

The image data stored in the frame memory 15 will be described in detail later in <Contents of the frame memory 15 of the decoding unit 10>.
<Encoding unit 20>
The encoding unit 20 includes a subtraction unit 21, an orthogonal transformation unit 22, a quantization unit 23, an inverse quantization unit 24, an inverse orthogonal transformation unit 25, an addition unit 26, a frame memory 27, a motion compensation prediction unit 28, and a variable length coding. The unit 29 is configured.

The subtracting unit 21 outputs a residual signal as it is, or by subtracting the pixel data of the block input from the adding unit 14 from the prediction signal from the motion compensation prediction unit 28.
The orthogonal transform unit 22 and the quantization unit 23 perform orthogonal transform and quantization on the block data from the subtraction unit 21 and output the result.
The inverse quantization unit 24, the inverse orthogonal transform unit 25, and the addition unit 27 restore the pixel data of the block by the inverse orthogonal transform, the inverse quantization, and the addition of the prediction signal, and output to the frame memory 27.

The frame memory 27 stores and holds a reference image made up of restored block pixel data.
The motion compensation prediction unit 28 obtains a prediction signal and a motion vector for the block subjected to motion compensation prediction, outputs the prediction signal to the subtraction unit 21 and the addition unit 26, and outputs the motion vector to the variable length coding unit 29. To do.

The processing method used when the motion compensation prediction unit 28 obtains a motion vector is generally a method for obtaining a motion vector from a reference image by motion search.
However, in the present invention, the motion vector is also obtained by a method for obtaining the motion vector output from the motion vector construction unit 30 as it is or by searching for a peripheral image based on the motion vector from the reference image. .

The variable length encoding unit 29 performs variable length encoding on the block data output from the quantization unit 23, the motion vector output from the motion compensated prediction unit 28, and the macroblock type, and encodes the resulting MPEG-2 format. Output a column.
<Contents of Frame Memory 15 of Decoding Unit 10>
Next, reference images stored in the frame memory 15 of the decoding unit 10 will be described with reference to FIGS. 27 and 28. FIG.

In this embodiment, the motion vector configuration unit 30 determines whether or not the motion vector is reused. The frame memory 15 stores information necessary for the determination.
FIG. 27 is a diagram showing the relationship of images stored in the frame memory 15.
FIG. 27A is a diagram showing a plurality of pictures included in a GOP, FIG. 27B is a diagram showing the display order of pictures, and FIG. 27C is a diagram showing a frame memory. is there.

The pictures in the GOP are arranged in the order in which they can be referred to during decoding, not in the display order. For example, since “B5” has “I1” and “B3” as reference pictures, “I1” and “B3” are positioned before “B5” (see FIG. 27A), but the display is “B5”. "Is before" B3 "(see FIG. 27B).
Decoding is performed in the order of pictures in the GOP, and a predetermined number of pictures are stored in the frame memory.

For example, when the predetermined number is 6, six pictures are stored in the frame memory, and the display order is managed. When “B5” is the target picture, reference picture numbers are assigned to other pictures in ascending order (see FIG. 27B).
The pictures stored in the frame memory are temporarily stored until they are displayed in the “reference picture”, which is a picture that is referenced from the subsequent picture, and in the order in which they should not be referenced but displayed in the subsequent picture. Are classified as “non-reference pictures”. For example, B5 is a non-reference picture, and “I1” and “B3” are reference pictures (see FIG. 27C).

When the reference picture is no longer referred to, it becomes a non-reference picture and is sequentially displayed in the order to be displayed.
The picture here is composed of a plurality of decoded macroblocks as shown in FIG. 24, and stores management data in addition to these image data.
FIG. 28 is a diagram illustrating an example of reference picture management data stored in the frame memory.

FIG. 28A is a diagram illustrating an example of the reference picture list 2800.
This reference picture list 2800 manages reference pictures stored in the frame memory, and is stored in the frame memory. The contents of the reference picture list 2800 are updated when a picture is displayed or a new picture is stored.

The reference picture list 2800 includes a display order 2801, a picture type 2802, and a storage address 2803.
The display order 2801 is an identification number for the target block to specify the reference picture 2321. For example, in FIG. 27B, the reference picture 2321 of the macroblock included in “B5” is “1” for “ref.idx0” and “4” for “ref.idx1” (see FIG. 23). .

This reference number is assumed to be assigned in ascending order in the display order with the target picture as a reference, and it is assumed that the reference number is assigned according to a predetermined rule that indicates whether the picture is in front of or behind the target block. This method is not limited to the method shown in FIG.
A picture type 2802 represents the coding type of this picture, and stores the picture type of picture data (see the third row in FIG. 23).

An address 2803 represents an address in the frame memory in which image data (see FIG. 24) is stored.
FIG. 28B is a diagram showing an example of the macroblock address list 2820.
A macro block position 2821 for specifying a macro block and an address 2822 in the frame memory 15 in which image data of the block is stored are configured. Normally, picture image data is stored in a continuous area. Therefore, if there is a storage address 2803, the address of each macroblock can be obtained. However, in this embodiment, for convenience of explanation, a macroblock address list 2820 is used. Suppose you have

The contents of the frame memory 27 of the encoding unit 20 are almost the same. The difference is that the picture type 2802 is not stored. This is because the encoding is unnecessary.
<Motion vector configuration unit 30>
The motion vector configuration unit 30 includes a determination unit 31, a program storage unit 32, and a prediction control unit 33.

The program storage unit 32 stores a program for determining a processing method of the motion compensation prediction unit 28 in advance.
The determination unit 31 determines a processing method to be performed by the motion compensation prediction unit 28 by executing the program stored in the program storage unit 32, and gives an instruction according to the determined processing method to the prediction control unit 33. Let it be done. The prediction control unit 33 gives an instruction to the motion compensation prediction unit 28 according to the determined processing method.

More specifically, the determination unit 31 receives the motion information 2320 from the variable length decoding unit 11, and includes the motion vector 2322 (see FIG. 23) and the picture type 2802 of the reference picture 2321 (see FIG. 28). Based on the above, new motion information is created, and the encoding unit 10 performs an encoding process using the new motion information.
The picture type 2802 acquires the picture type of the corresponding picture from what was sent from the variable length decoding unit 11 and once stored in the frame memory 15.

Hereinafter, the processing of the motion vector construction unit 30 will be described in detail.
2 to 4 are flowcharts showing the procedure of the program stored in the program storage unit 32. Symbols A to E in the respective drawings are connectors, and connectors having the same symbol in the drawings are combined to represent one flowchart.
The contents of the program are simply as follows: Step (1): The relationship between the macro block in the MPEG-4 / AVC format moving image and the corresponding reference image is specified. Step (2): The relationship is It is determined whether or not the condition regarding the relationship defined in the MPEG-2 format is met, and step (3): a processing method to be performed by the motion compensation prediction unit 28 is determined according to the determination result. is there. As for the correspondence with the drawing, (1) corresponds to the whole of FIG. 2, (2) corresponds to steps S301 to S304 in FIG. 3 and steps S401 to S404 in FIG. 4, and (3) corresponds to FIG. This corresponds to steps S305 to S312 and steps S405 to S410 in FIG.

  An AU delimiter (Access Unit Delimiter) is attached to the head of a picture and indicates the head position. This AU delimiter includes primary_pic_type which is information indicating the slice type of the slice included in the picture. Using this, when the AU delimiter is present in the variable-length decoded data, the determination unit 31 acquires primary_pic_type included therein and identifies the picture type from the slice type indicated by primary_pic_type.

  The slice type limits the coding mode of the macroblock in the slice. More specifically, the P slice is a slice for performing intra-picture coding and inter-picture predictive coding using one reference picture. The B slice is a slice for performing intra-picture coding and inter-picture predictive coding using one or two reference pictures. The I slice is a slice that performs only intra-frame coding. In MPEG-4 / AVC, a slice type indicating the type of slice is attached to each slice, and a single picture or a plurality of slice types are included in one picture. Therefore, if all the slices in the picture are the same slice type, the determination unit 31 sets the slice type as the picture type. For example, if all slices are P slices, the picture is specified as a P picture, and if all slices are I slices, the picture is specified as an I picture.

  When a plurality of slices are mixed in the picture, the determination unit 31 identifies the B picture when the B slice is included, and determines that the P slice is included when the B slice is not included. Identify as a picture. For example, when an I, P, or B slice is included in a picture, the determination unit 31 specifies a B picture because the B slice is included. When the I and P slices are included and the B slice is not included, the P picture is specified by the P slice being included.

  The specification of the picture type is not limited to the above method. For example, the determination unit 31 may specify the picture type from the slice type acquired by acquiring the slice type from each slice in the picture among various data separated by the variable length decoding unit 11. At this time, it is not necessary to acquire all the slice types in the picture, and if there is even one B slice in the process of sequentially acquiring the slice types in the picture, the B picture may be specified.

  When the picture type is specified in this way, next, the determination unit 31 obtains the macroblock 2300 (see FIG. 23) of the target block from the encoded information separated by variable length decoding in steps S202, S205, and S206. Then, the macro block type 2310 or the sub macro block type is acquired and the prediction mode indicated in them is determined.

When the picture type is determined at the head of the picture, the picture type is not specified until an AU delimiter appears thereafter.
First, it is determined whether the target block is a P picture (step S201) or a B picture (step S204), and the picture type “I / P / B” of the picture including the target block is allocated. .

If it is an I picture (step S204: NO), there is no motion vector to be reused, and the process moves to a process not using a motion vector (D).
If it is a P picture (step S201: YES), it is determined whether or not the prediction mode is inter-screen prediction (step S202). If it is inter-screen prediction (step S202: YES), the frame memory 15 is used. A reference image of a motion vector is acquired (step S203).

Here, the reference image is acquired by acquiring screen data from the storage address 2803 (see FIG. 28) of the display order 2801 corresponding to the reference picture 2321 of the macroblock 2300. At the same time, a reference picture list 2800 is also acquired.
If the prediction is in-screen prediction or non-reference (step S202: NO), there is no motion vector to be reused, and the process moves to a process not using a motion vector (D).

If it is a B picture (step S204: YES), it is determined whether or not the prediction mode is inter-screen prediction (step S205). (Reference to two directions) (step S206) is determined, and reference images of motion vectors are respectively acquired (step S207 and step S208).
Hereinafter, each picture type will be described.

(P picture)
First, the case of P picture (A) will be described (see FIG. 3).
In step S301, the determination unit 31 determines whether the relationship between the target block of the P picture and the reference image conforms to the MPEG-2 format.
Here, the MPEG-2 standard for P picture blocks and reference pictures is that the reference picture is the past P picture or I picture that is closest in time to the picture of the target block.

In other words, the reference picture list 2800 acquired from the frame memory in step S203 is referred to, and if the picture type 2802 of the corresponding reference image is “P / I”, it is determined that it is suitable.
If all the picture types 2802 having numbers smaller than the display order 2801 of the corresponding reference image are all “B”, it is determined that they are closest in time. For example, in FIG. 27B, when the reference picture number of the reference image is “1”, the picture type of the reference picture number “0”, that is, the picture type 2802 whose display order 2801 is “3” is “B”. If there is, it is judged to be the closest in time.

The method for determining the matching of the picture type with reference to the picture type 2802 and the method for determining the closest in time from the display order 2803 are the same in the following matching determination.
As a result of the determination, if it matches, it is determined that reference is possible (step S302: YES), and the process proceeds to step S305 in order to use the motion vector or the like for MPEG-2 encoding.

If it is determined that the reference is impossible without matching (step S302: NO), the process proceeds to a process not using a motion vector (step 309).
(B picture: see 1 direction)
Next, the case (B) of referring to one direction in a B picture will be described (see FIG. 3).
In step S303, the determination unit 31 determines whether or not the relationship between the target block of the B picture and the one-way reference image meets the MPEG-2 condition. Here, the MPEG-2 standard for a macroblock of a B picture and one reference picture is that the reference picture is the past P picture or I picture that is closest in time to the picture of the target block. is there.

As a result of the determination, if it matches, it is determined that it can be referred to (step S304: YES), and the process proceeds to step S307 for use in MPEG-2 encoding such as a motion vector.
If it is determined that reference is impossible without matching (step S304: NO), the process proceeds to a process not using a motion vector (step 309).
(B picture: see 2 directions)
Next, the case (C) in which the B picture is referred to in two directions will be described (see FIG. 4).

  In step S <b> 401, the determination unit 31 determines whether the relationship between the target block of the B picture and the reference image for two-way reference meets the MPEG-2 condition. Here, the definition of MPEG-2 for the block of B picture and the two reference images is “the past P picture in which one of the two reference images is closest in time order from the picture of the target block” Alternatively, it is an I picture, and the other reference picture is a future P picture or I picture that is closest in time order from the picture of the target block.

In this determination, the determination unit 31 classifies the determination results into four types of steps S402, S403, S404 and others (E) as described below.
First, only one reference image is the closest past or future I / P picture.
The second is the closest past or future I / or P picture for both of the two reference images.

The third is the closest past I / P picture, and the one is the closest future I / P picture.
The fourth is the other case, that is, the case where the two reference images are not compatible.
If it is determined in step S402 that one of the two reference images conforms to the MPEG-2 standard, the process proceeds to step S405.

In step S403, if both of the two reference images conform to the MPEG-2 standard and both are forward reference or both are backward reference, the determination unit 31 advances the process to step S407.
In step S404, if both of the two reference images conform to the MPEG-2 standard, and one reference image is a forward reference and the other reference image is a backward reference, the determination unit 31 proceeds to step S409. Proceed with the process.

In other cases (step S404: NO), the process proceeds to (E).
(Step (3))
Finally, the determination unit determines a processing method to be performed by the motion compensation prediction unit 28 according to the determination result in step (2), and the prediction control unit 33 causes the motion compensation prediction unit 28 to perform a processing method according to the determination. Instruct.

Hereinafter, each picture type will be described.
(P picture)
When the reference image for the target block of the P picture in MPEG-4 / AVC is the past P picture or I picture closest in time order to the picture of the target block, the determination unit 31 sends the motion compensation prediction unit 28 Then, it is determined to reuse the motion vector (step S305). In other words, the motion compensation prediction unit 28 encodes an MPEG-4 / AVC motion vector as it is as an MPEG-2 motion vector, instead of searching for a motion vector from a reference image by motion search. Means that.

In this case, the determination unit 31 determines that the MPEG-4 / AVC macroblock type (inter-screen reference) is also maintained in MPEG-2 (step S306, see FIG. 5).
(B picture: see 1 direction)
When one reference image for a block of B picture in MPEG-4 / AVC is a past or future P picture or I picture closest in time order to the picture of the target block, the determination unit 31 The motion compensation prediction unit 28 is determined to reuse the motion vector (step S307).

In this case, the determination unit 31 determines that the MPEG-4 / AVC macroblock type (one-way reference) is also maintained in MPEG-2 (see step S308, FIG. 5).
(B picture: see 2 directions)
In addition, when only one of two reference images for a block of a B picture in MPEG-4 / AVC is a past or future P picture or I picture closest in time to the picture of the target block. The determination unit 31 determines that the motion compensation prediction unit 28 should reuse the motion vector (step S405).

In this case, the determination unit 31 determines to make the macroblock type in MPEG-2 refer to one direction (step S406, see FIG. 9).
Also, in the MPEG-4 / AVC, the two reference images for the B picture block are both the past or future P picture or I picture closest in time order to the picture of the target block, and the reference direction If both are forward references or both are backward references, the determination unit 31 selects one of the motion vectors and decides to reuse it (step S407).

The selection of the motion vector may be configured to select, for example, a smaller value of the residual signal of the corresponding block, or may be configured to select a smaller motion vector.
In this case, the determination unit 31 determines that the macroblock type in MPEG-2 is referred to as one direction (step S408, see FIG. 10).
Also, two reference pictures for a B picture block in MPEG-4 / AVC are both past or future P pictures or I pictures that are closest in time to the picture of the target block, and one of them is If the other is a forward reference and the other is a backward reference, the determination unit 31 determines to reuse the two motion vectors (step S409).

In this case, the determination unit 31 determines to maintain the macroblock type (refer to both directions) in MPEG-4 / AVC also in MPEG-2 (see step S410, FIG. 7).
When the reference image for the target block of the B picture does not conform to the MPEG-2 standard, the determination unit 31 determines not to use the motion vector of the target block, and sets the macroblock type in MPEG-2. No inter-screen reference is made (see steps S309 and S310, FIG. 6 and FIG. 8).

(I picture)
For the target block of I picture and the target block for non-reference and intra prediction in P / B picture, the determination unit 31 determines not to use a motion vector (step S311), and MPEG-4 / AVC It is determined that the macro block type (no inter-screen reference) is maintained even in MPEG-2 (step S312).

In response to the determination by the determination unit 31, the prediction control unit 33 sends a motion vector, a macroblock type, information indicating which image is a reference image, and the like to the motion compensation prediction unit 28, and uses them to code To instruct
Specifically, when the motion vector can be reused, the motion information 2320 is transmitted to the motion compensation prediction unit 28, and when the motion vector cannot be reused, the fact is notified.
(Operation example)
Next, the operation will be described with reference to an example of a relationship between the macroblock and the reference image.

FIG. (A) shows an input image when a moving image is converted by the conversion apparatus according to the present invention, and (b) shows an output image.
In both FIG. 1 (a) and FIG. 1 (b), the same moving images are arranged in order of display from left to right. Fig. (A) shows the relationship between the target block and the reference block in the MPEG-4 / AVC format, and Fig. (B) shows the MPEG-2 after conversion for the same target block as Fig. (A). The relationship with the reference image in the format is shown. The same applies to FIGS.

First, typical patterns will be described with reference to FIGS.
FIG. 5 shows a case of using a motion vector in the case of referring to one direction.
This case corresponds to the case of the P picture and the case of referring to the B picture in one direction.
In FIG. 5A, the target block of the right P picture is (step S201; YES), and the reference block of the left I picture is used as a reference image by forward inter-frame prediction (step S202; YES). This reference relationship conforms to the definition of MPEG-2 (step S302; YES). The motion vector at this time is MV1.

In this case, the motion compensation prediction unit 28 of the encoding unit 20 applies the same motion vector as that of FIG. 5A to the target block of the P picture according to the instruction from the prediction control unit 33, as shown in FIG. Using MV1 as it is, it is encoded using a motion vector obtained by motion search, and the macroblock type is used as an inter-screen reference.
FIG. 6 shows a case where a motion vector cannot be used, accompanied by a change in the macroblock type.

This case corresponds to the case of the P picture and the case of referring to the B picture in one direction.
FIG. 7 shows a case of using a motion vector in the case of bi-directional reference.
This case corresponds to the case of bi-directional reference of a B picture.
In FIG. 7A, the picture of the target block is a B picture (step S204; YES), and bi-directional reference inter-screen prediction is performed (step S205; YES, step S206; NO). Reference can be made in the MPEG-2 standard (step S404; YES). Therefore, as shown in FIG. 7B, the two motion vectors MV1 and MV2 are used (step S409), and the macroblock type is referred to as bidirectional (step S410).

FIG. 8 shows a case where two motion vectors cannot be used in the case of bi-directional reference.
This case corresponds to the case of bi-directional reference of a B picture.
In FIG. 8A, the target block in the B picture refers to two B pictures, but in MPEG-2, the B picture cannot be used as a reference image (steps S204 → S205 → S206 → S208 → S401). → S402 → S403 → S404; NO).

Therefore, as shown in FIG. 7B, the motion compensation prediction unit 28 does not use the motion vector in FIG. 5A and encodes the macroblock type without reference between screens (steps S309 and S310).
FIG. 9 is a case of using a motion vector accompanied by a change of the macroblock type.
This case corresponds to the case of bi-directional reference of a B picture.

In FIG. 9A, the target block in the B picture refers to the I picture and the B picture. In MPEG-2, reference to an I picture is permitted, but reference to a B picture is not permitted (steps S204 → S205 → S206 → S208 → S401 → S402; YES).
Therefore, the motion compensation prediction unit 28 uses the motion vector MV1 that refers to the I picture and sets the macroblock type as the inter-screen reference in one direction (steps S305 and S306), as shown in FIG.

FIG. 10 shows a case where some motion vectors cannot be used, accompanied by a change in macroblock type.
This case corresponds to the case of bi-directional reference of a B picture.
In FIG. 10A, the target block in the B picture refers to the two reference blocks 1 and 2 in the front I picture. In MPEG-2, it is not permitted to refer to the two images in the front or to refer to the two images in the back (step S204 → S205 → S206 → S208 → S401 → S402 → S403; YES). .

Therefore, as shown in FIG. 10B, the motion compensation prediction unit 28 selects and reuses MV1, which is one of the two motion vectors MV1, MV2, and changes the macroblock type from the previous two-direction reference to the previous one. Encoding is performed instead of the direction reference (steps S407 and S408).
<Modification 1>
In the first embodiment, a typical pattern has been described with respect to motion vector reuse. However, even when motion vector reuse is possible, it cannot be used as it is or when it is better not to use it as it is. There is.

An example in which the former cannot be used as it is will be described with reference to FIG. 11, and an example in which the latter should not be used as it is, will be described with reference to FIG.
First, when it cannot be used as it is, there is a problem of pixel accuracy.
That is, the pixel accuracy is 1/4 in MPEG-4 / AVC and 1/2 in MPEG-2.

FIG. 11A shows the same situation as FIG. 9A, in which only one of the two vectors of the B picture is reused. However, the present invention is not limited to this example.
Here, the pixel accuracy of the motion vector is 1/4 in MPEG-4 / AVC, and the pixel accuracy is 1/2 in MPEG-2. Therefore, as shown in FIG. 11 (b), the motion compensation prediction unit 28 rounds the motion vector MV1 to ½ precision when the motion vector MV1 in FIG. 11 (a) is specified in units of ¼ pixels. Correction is performed, and encoding is performed using the motion vector MV3 obtained as a result. This process is performed by the motion vector construction unit 30 and a new motion vector is transmitted to the motion compensation prediction unit 28. Alternatively, the motion compensation prediction unit 28 may perform this.

Next, as a case where it is better not to use it as it is, there is a case where the macroblock type is changed.
FIG. 12A shows the same situation as FIG. 9A, in which only one of the two vectors of the B picture is reused, and the macroblock type is changed.
In the case of bi-directional reference, an optimal predicted image is obtained by taking the average of two reference blocks (reference block 1 and reference block 2). That is, an inter-screen prediction error 2330 (see FIG. 23) having a small data amount can be obtained.

However, by using one reference block, the predicted image changes, and the error from the predicted image may be larger than expected.
If there is a possibility that the error becomes large, it is better to search the motion vector again to obtain an optimal new motion vector, but there is a drawback that the processing amount of the motion vector search is large.
Therefore, by setting only the vicinity of the reference block of the motion vector (MV1) to be reused as the re-search range, an optimal new motion vector can be obtained and the processing time can be greatly shortened.

As shown in FIG. 12B, another motion vector MV3 is detected by re-searching a predetermined range based on the position of the reference block 1 pointed to by the MV1 on the reference image, and the motion vector MV3 is obtained. Use and encode.
More specifically, the motion compensation prediction unit 28 expands a search range of, for example, ± 30 pixels in the x and y directions with the coordinate position of the reference block 1 as the center, and a block having a smaller prediction error due to motion search from the search range. And the motion vector pointing to that block is MV3.

The motion vector construction unit 30 designates the reference image and the reference block, and gives an instruction to the motion compensation prediction unit 28 to re-search the neighborhood of the reference block.
As a method of obtaining the reference block data included in the motion vector search range, for example, the motion vector construction unit 30 can obtain the data as follows.
FIG. 29 is a diagram illustrating a target block and a macroblock in the vicinity thereof. Here, for example, the data of the macroblocks included in the search range 1 are eight such as (2, 2) and (2, 3) and the target block (3, 3).

The macro block data can be easily obtained by referring to the macro block address list 2820 of the frame memory 27. For example, if the macroblock position 2821 is “(0, 2)”, data for one block may be read from the address 2822 “addr02”.
Similarly, in the case of the search range 2, the macro block (1, 1) including the range is read.

<Modification 2>
Here, even when it is determined that the motion vector cannot be reused in the first embodiment, a case where the reuse can be performed by changing the reference block of the motion vector will be described.
Although it is conceivable that the prediction error increases by changing the reference block, there is an advantage that it is not necessary to perform a re-search process of the motion vector.

FIG. 13A shows the same situation as FIG.
In FIG. 9B, the motion vector “MV2” is discarded and only the motion vector “MV1” is used for encoding.
On the other hand, as shown in FIG. 13B, encoding may be performed using the motion vector “MV3” instead of the motion vector “MV2”.

The motion vector “MV3” has the same value indicating the same spatial coordinates as the motion vector “MV2”, but the point that the pointing reference image is an I picture or P picture temporally close to the original reference image is different.
This is because images that are close in time are likely to have similar pixel values.
The motion compensation prediction unit 28 selects the reference image 3 from images whose relationship with the target block conforms to the definition of MPEG-2. This process may be performed by the motion vector configuration unit 30.

Since the reference block 3 may have a large error with the target block, the reference block 3 may be configured to search for a predetermined range around the reference block 3 and find a reference block with a small error (see below). 21).
<Embodiment 2>
In the first embodiment, an example of a frame has been dealt with, but it may be applied to a field structure as shown in FIGS.

The field is obtained by dividing one frame (screen) into two, a Top field that is a set of odd-numbered scan lines and a Bottom field that is a set of even-numbered scan lines. .
For example, in a normal television, a method of alternately displaying a Top field and a Bottom field is adopted, which is called an interlace method.

In the case of such a screen structure, it is possible to perform encoding by handling one field like one frame.
That is, it is possible to replace the field with a frame and convert the data format as described in the first embodiment, that is, convert the moving image data created by the high compression encoding method into MPEG-2 data. Become.

Here, looking at the specifications of the MPEG-2 standard, it is possible to refer to both fields of the reference image from both the own field for both the P picture and the I picture. In addition, in the case of an I picture, the Macro field in the Bottom field can refer to the Top field.
More specifically, in the MPEG-2 field structure, the reference image is recognized as a reference image when the picture to be encoded is an I picture, the Top field cannot be referred to, and the top field can be referred to only in the Bottom field. In the case of a P picture, both the Top field and the Bottom field can refer to the past P picture field or the I picture field that is closest in time order. In the case of a B picture, one is a time field in both the Top field and the Bottom field. It is possible to refer to both fields of the past P picture or I picture that are closest in order, and one can refer to both fields of the future P picture or I picture that are closest in time order. This restriction is the same in MPEG-4.

Therefore, in a standard that enables frame reference of a reference image with the same restrictions as the current MPEG-4 / AVC, conversion similar to that in the first embodiment is possible.
For example, in the example of FIG. 14, the reference images are “bottom field of I picture” and “top field of B picture”, and only the former can be referred to as the MPEG-2 reference image. Therefore, by changing the macroblock type to one-way reference, the motion vector can be reused with the former as a reference image.

Currently, the MPEG-2 standard does not support inter-frame references as freely as in the MPEG-4 / AVC standard, but may be supported in the future.
This is because the moving picture coding schemes appearing after MPEG-2 and MPEG-4 tend to increase the degree of freedom of reference images in order to increase the compression rate, as in MPEG-4 / AVC. Because it is done.

As one of the expected encoding methods, in addition to the current MPEG-2 standard, the Bottom field of a B picture can refer to the Top field of the same frame, but other MPEG pictures cannot be referred to as in MPEG2. A method is conceivable.
That is, this method can be said to be a moving image encoding method that has a higher degree of freedom than the restriction of the MPEG-2 reference image and a lower degree of freedom than the restriction of the MPEG-4 / AVC.

Like MPEG-4 / AVC, if reference image candidates can be up to 16 frames (32 pictures), it is necessary to leave the reference image in memory at the time of decoding. This is because a load is applied.
FIG. 15 shows a method of conversion from MPEG-4 / AVC to a moving picture coding system having such a reference picture restriction.

The example in FIG. 15 is an example in which only the field reference in the same frame is reused without using the motion vector because the reference image is a B picture.
In the encoding system A having restrictions such as MPEG-2, since the target screen is a B picture, any motion vector cannot be reused, but “the bottom field of a B picture can refer to the Top field of the same frame. In the encoding method B having the constraint “”, the motion vector of the field reference in the same frame can be reused.

In FIG. 15, as one of the expected encoding formats, the description has been made assuming an encoding method in which reference is possible within the same frame in addition to the restriction of the reference image of MPEG-2. The same explanation holds true for any coding scheme having any reference image.
For example, the same may be considered when converting a moving image encoded in one of the above-described expected encoding formats (encoding method B) to MPEG-2. As an example of this case, as shown in FIG. 16, only one of the two motion vectors can be reused. This is because MPEG-2 cannot refer to the B picture within the same frame. FIG. 16A shows a motion vector in the encoding method B, and FIG. 16B shows a motion vector in MPEG-2.
<Embodiment 3>
In the first embodiment, the reuse of motion vectors present in MPEG-4 / AVC format data has been described. In the present embodiment, although not present in the data, it is dynamically obtained when decoding a macroblock. A case where a motion vector to be reused will be described.

In MPEG-4 / AVC, a direct mode is defined in which a reference image can be referred to without a motion vector.
The direct mode is provided in order to increase the compression efficiency, and is a coding mode for predicting and generating motion information from motion information of other blocks. That is, the amount of data relating to motion information can be reduced.

In MPEG-2, the direct mode is not supported, and in the conversion from MPEG-4 / AVC, there is no motion vector to be reused, so that the reference image cannot be referred to.
In this embodiment, a virtual motion vector calculated at the time of MPEG-4 / AVC decoding is used as a motion vector of MPEG-2. That is, a motion vector that did not exist in MPEG-4 / AVC is calculated and added in MPEG-2.

There are two types of direct mode, which will be described with reference to FIGS.
The first direct mode motion information generation will be described with reference to FIG.
The first direct mode is a spatial direct mode in which a motion vector of a target macroblock is virtually obtained using motion vector information of macroblocks at peripheral positions.

In the spatial direct mode, prediction is basically performed using three macroblocks, a macroblock 1701 on the left side of the target block, an upper macroblock 1702 and a macroblock 1703 on the upper right.
Among the motion vectors of these three macroblocks (1701, 1702, 1703), the prediction method uses a motion vector indicating a past reference image that is closest in time and a future motion vector that is closest in time. Select and predict as a virtual motion vector of the target block. The target block is a two-way reference.

In FIG. 17A, the motion vectors of the three macroblocks are “MV5” for the macroblock 1701, “MV1” for the macroblock 1702, and “MV2” for the macroblock 1703, respectively.
Among them, the motion vector “MV1” indicating the past reference image closest in time and the future motion vector “MV2” closest in time are selected and used as the motion vector of the target block as they are.

For example, the motion vectors of the target block are “MV3” and “MV4” (see FIG. 17B).
“MV3” has the same vector value as “MV1”, and “MV4” is the same as “MV2”. The reference images are the same reference image 1 and reference image 2 as “MV1” and “MV2”. The reference block 1 and the reference block 2 are determined from the reference image and the motion vector.

Next, generation of motion information in the second direct mode will be described with reference to FIG.
The second direct mode is a temporal direct mode in which the motion vector of the target block is virtually obtained using the motion information of the macro block at the peripheral time.
In the temporal direct mode, basically, a motion vector for a target macroblock is estimated from a motion vector indicating a macroblock 1801 at the same position as the target block in a future reference image (reference image 2) that is closest in time. Thus, the virtual motion vector of the target block is obtained. The target block is a two-way reference.

The virtual motion vector is obtained by calculating the motion vector “MV1” of the macroblock 1801 from the ratio of the time interval between the reference image 1 and the reference image 2 and the time interval between the reference image 1 and the target image “MV2” and “MV3”. Ask for.
The obtained “MV2” and “MV3” are used as motion vectors of the target block. In this case, the reference block “MV2” becomes the reference block 1 of the same reference image 1 as the macroblock 1802, and the reference block “MV3” becomes the reference block 2 of the same reference image 2 as the macroblock 1801.
<Embodiment 4>
In the present embodiment, when reusing a motion vector, the motion vector reference image type “I / P” and the time direction requirement are satisfied, but the reference block position requirement is satisfied. The case where there is not will be described with reference to FIGS. 19 and 20.

In MPEG-4 / AVC, the concept of reference images is extended to the outside of the screen, so that a motion vector may refer to a macroblock that protrudes outside the screen. For example, the macro block 1901 of the reference image 1 in FIG.
The area outside the screen is referred to assuming that the outside of the screen is filled with the endmost pixel.
On the other hand, in MPEG-2, it is necessary that the motion vector always refers only to the screen.

For example, if the target block is a two-way reference block and only the motion vector “MV1” refers to the outside of the screen as shown in FIG. Only the motion vector “MV2” can be reused, but in this example, it cannot be reused because it is a backward reference.
Also, even when all motion vectors refer to the outside of the screen, they cannot be reused.

Therefore, as shown in FIG. 20, it is possible to refer to MPEG-2 by correcting the motion vector so that the reference outside the screen becomes the reference inside the screen.
In this case, MPEG-2 uses a reference block (reference block 2) that is different from the reference block (reference block 1) that was initially assumed, so that an error from the predicted image may occur.

Assuming such a case, as described in the first embodiment, a process may be performed so as to reduce the generation of errors from the predicted image as much as possible by performing a re-search of the region or the like ( (See FIG. 12).
<Supplement>
As mentioned above, although embodiment of this invention was described, this invention may be as follows not only the said form.
(1) In the embodiment, basically, a reference image including a reference block is reused without changing the motion vector, but the motion vector is reused by changing the reference image. It is good. The case described here can be said to be an extension of the case described with reference to FIG.

An example is shown in FIG.
In FIG. 21A, since the reference image (reference image 1) is a B picture, it cannot be reused. However, by using the P picture as the reference image (reference image 2), the motion vector “MV1” may be reusable.
However, in this case, since an error from the predicted image may increase, the vicinity of the reference block of the P picture is re-searched as a search range, and a new reference block (reference block 2) and a new motion vector “MV2” are obtained. It is good also (refer FIG.21 (b)).

In the example of FIG. 21, only the reference image is changed, but the position of the reference block may be changed together with the reference image without re-searching.
An example is shown in FIG.
In FIG. 22B, the reference image is changed to a P picture (reference image 2), and the motion vector “MV1” is set to the time interval between reference image 1 and reference image 2, and the time interval between reference image 1 and target image. From this ratio, the position where the reference block is moved is determined, and a new reference block (reference block 2) is obtained.

Also in this case, since an error with the predicted image may increase, a new motion vector “MV2” may be obtained by re-searching the vicinity of the reference block of the P picture as a search range.
(2) Although the embodiment has been described on the assumption that MPEG-4 / AVC is converted to MPEG-2, the present invention is not limited to these encoding methods, and restrictions on motion vectors. Any encoding method may be used as long as the conversion is between different encoding methods.

  For example, conversion from MPEG-4 / AVC to MPEG-4, conversion from MPEG-4 / AVC to VC1, conversion from MPEG-4 / AVC to AVS (Audio Video coding Standard of China), MPEG-4 / AVC Conversion from the main profile to the MPEG-4 / AVC baseline profile may be performed, or conversion between other encoding methods may be performed.

In the embodiment, picture-to-picture conversion is described. However, slice-to-slice and picture-to-slice conversion may be used.
(3) Although the conversion from MPEG-4 / AVC to MPEG-2 has been described in the embodiment, for example, conversion from MPEG-2 to MPEG-4 / AVC may be performed.
As a difference between MPEG-2 and MPEG-4 / AVC, there is a difference in motion vector search range. The maximum search range of motion vectors defined in MPEG-4 / AVC is smaller than the maximum search range of MPEG-2.

When converting from MPEG-2 to MPEG-4 / AVC, the moving image conversion apparatus determines whether the motion vector for the MPEG-2 block falls within the maximum search range defined in MPEG-4 / AVC. It may be configured to determine whether to use a motion vector depending on whether or not.
At this time, for a motion vector whose motion vector does not fall within the maximum search range defined in MPEG-4 / AVC, the motion vector is reused by shifting the motion vector so that it falls within the maximum search range. You may comprise.
(4) The moving image conversion apparatus 100 may realize all or a part of each component in FIG. 1 with an integrated circuit of one chip or a plurality of chips, or with a computer program, Any other form may be used.

The determination unit 31 is configured to determine a processing method to be performed by the motion compensation prediction unit 28 by executing a program stored in the program storage unit 32. Instead of this configuration, a logic circuit is used. You may comprise. The logic circuit uses, as input signals, information necessary for determination, such as information on slice types, macroblock types, and reference images, from various data separated by the variable length decoding unit 11, and performs motion compensation prediction according to the input signals. An output signal indicating a processing method to be performed by the unit 28 is output.
(5) In the case of a computer program, a program written in any recording medium such as a memory card or CD-ROM may be read and executed by a computer, or the program is downloaded and executed via a network. You may make it form.
(6) In the above embodiment, an example is described in which moving image data in units of slices is converted into moving image data in units of pictures. However, the present invention is also applicable to conversion from slice units to slice units and from picture units to picture units. The concept of can be applied.
(7) The moving image conversion apparatus 100 may be configured by an LSI with all or some of the components in FIG.

  The present invention can be applied to AV equipment that performs moving image conversion processing.

It is a block diagram which shows the structure of a moving image converter. 4 is a flowchart (part 1) illustrating a procedure of a program stored in a program storage unit 32. 4 is a flowchart (part 2) illustrating a procedure of a program stored in a program storage unit 32. 7 is a flowchart (No. 3) showing a procedure of a program stored in a program storage unit 32. It is a figure which shows the operation example in the case of hold | maintaining 1 direction reference. It is a figure which shows the operation example in case 1 direction reference is made into the non-screen reference. It is a figure which shows the operation example in the case of hold | maintaining a two-way reference. It is a figure which shows the operation example in case 2 direction reference is made into the non-screen reference. It is a figure which shows the operation example in case 2 direction reference (front, back) is made into 1 direction reference. It is a figure which shows the operation example in case 2 direction reference (front, front) is made into 1 direction reference. It is a figure which shows the operation example in the case of converting from pixel accuracy 1/4 to 1/2. It is a figure which shows the operation example in the case of a motion vector re-search. It is a figure which shows the operation example when making a reference image into an image from which a picture type differs. It is a figure which shows the operation example in the case of the reference between fields. It is a figure which shows the operation example in the case of the reference between fields in the same flame | frame. It is a figure which shows the operation example in the case of the reference between fields. It is a figure which shows the operation example in making it refer to two directions from a spatial direct mode. It is a figure which shows the operation example in making it refer to two directions from temporal direct mode. It is a figure which shows the operation example in the case of referring an off-screen block. It is a figure which shows the operation example at the time of moving the reference block outside a screen into a screen. It is a figure which shows the operation example of the re-search when a reference screen is made into the screen from which a picture type differs. It is a figure which shows the operation example of the re-search when a reference block is moved to another screen. It is a figure showing the example of a structure of MPEG encoding data. It is a figure which shows the relationship between 1 screen and a macroblock. It is a figure which shows the example of the motion vector of MPEG-2. It is a figure which shows the example of the motion vector of MPEG-4 / AVC. It is a figure which shows the relationship of the picture in a frame memory. It is a figure which shows the example of a table which manages the picture in a frame memory. It is a figure which shows the example of a macroblock selection method at the time of a re-search.

Explanation of symbols

DESCRIPTION OF SYMBOLS 10 Decoding part 11 Variable length decoding part 12 Inverse quantization part 13 Inverse orthogonal transformation part 14 Addition part 15 Frame memory 16 Motion compensation part 20 Encoding part 21 Subtraction part 22 Orthogonal transformation part 23 Quantization part 24 Inverse quantization part 25 Inverse Orthogonal transformation unit 26 Addition unit 27 Frame memory 28 Motion compensation unit 30 Motion vector configuration unit 31 Determination unit 32 Program storage unit 33 Prediction control unit 100 Video image conversion device

Claims (19)

A moving image conversion apparatus that converts first moving image data encoded according to a first motion compensation prediction method into second moving image data having the same format as data encoded according to a second motion compensation prediction method. ,
Selecting means for selecting a motion vector satisfying a condition to be a motion vector in the second motion compensated prediction method of the block from one or more motion vectors of the block included in the first moving image data;
Encoding means for performing encoding according to a second motion compensation prediction method, comprising: encoding means for encoding a block using each motion vector selected by the selection means; A moving image conversion device. In the encoding using each selected motion vector, the encoding means performs encoding using each selected motion vector as it is, instead of using the motion vector specified by motion search. The moving image conversion apparatus according to claim 1, wherein: In the encoding using each selected motion vector, the encoding means is a predetermined search range based on the position on the reference image indicated by each motion vector, and the reference block indicated by the motion vector The moving image conversion apparatus according to claim 1, wherein a motion search is performed for a search range including and a block is encoded using a motion vector obtained as a result of the motion search instead of using the motion vector. . The condition to be satisfied by the motion vector selected by the selection means is to limit the encoding type of the reference image of the block,
The moving image conversion apparatus according to claim 1, wherein the selection unit selects a motion vector that satisfies the condition. The condition to be satisfied by the motion vector selected by the selection means is to limit the encoding type of the reference image of the block,
When the reference image of the motion vector of the block included in the first moving image data does not satisfy the condition, the selection means replaces the reference image with a reference image that satisfies the condition and selects the motion vector. The moving image conversion apparatus according to claim 1. In the encoding using each selected motion vector, the encoding means is a predetermined search range based on the position on the reference image indicated by each motion vector, and the reference block indicated by the motion vector 6. The moving image conversion apparatus according to claim 5, wherein a motion search is performed for a search range including the image and a block is encoded using a motion vector obtained as a result of the motion search instead of using the motion vector. . The condition to be satisfied by the motion vector selected by the selection means is that, when the motion vector of the block is one, the reference image of the motion vector is an image displayed before the image including the block. Yes,
The moving image conversion apparatus according to claim 1, wherein the selecting unit selects a motion vector when one of the motion vectors of the block included in the first moving image data satisfies the condition. The condition to be satisfied by the motion vector selected by the selection means is that when there are two motion vectors of a block, the reference image of one of the motion vectors is an image displayed before the image including the block. The reference image of the other motion vector is an image displayed after the image including the block,
The moving image conversion according to claim 1, wherein when two of the motion vectors of the blocks included in the first moving image data satisfy the condition, the selecting means selects the two motion vectors. apparatus. The condition to be satisfied by the motion vector selected by the selection means is that the coding type of the image displayed between the image including the block and the reference image of the block is a predetermined type in the image display order. And
The moving image conversion apparatus according to claim 1, wherein the selection unit selects a motion vector when a reference image of a motion vector of a block included in the first moving image data satisfies the condition. . The condition to be satisfied by the motion vector selected by the selecting means is that the reference image of the block is the image of the past P picture or I picture that is closest in display time order from the display of the image including the block, or the closest future Is an image of a P picture or I picture, or an image including the block,
The moving image conversion apparatus according to claim 1, wherein the selection unit selects a motion vector when a reference image of a motion vector of a block included in the first moving image data satisfies the condition. . The number of motion vectors of blocks included in the first moving image data is two,
When the selection means selects only one motion vector that satisfies the condition,
The encoding means encodes a block using the one motion vector, and obtains information indicating the type of motion prediction of the block from one type of motion prediction using two motion vectors. The moving image conversion apparatus according to claim 4, wherein the moving image conversion device is changed to a type of motion prediction to be used. The condition is that all the reference blocks of the block exist in the reference image included in the reference block,
The moving image conversion apparatus according to claim 1, wherein the selection unit selects a motion vector when a reference block of a motion vector of a block included in the first moving image data satisfies the condition. . If the reference block of the corrected motion vector is present in the reference image by correcting the motion vector among the motion vectors that have not been selected, the selecting means selects the corrected motion vector. The moving image conversion apparatus according to claim 12, wherein: The moving image conversion apparatus according to any one of claims 1 to 13, wherein the first motion compensation prediction method conforms to the MPEG4-AVC standard. The moving image conversion apparatus according to any one of claims 1 to 14, wherein the second motion compensation prediction method is based on the MPEG2 standard or the MPEG4 standard. A moving image conversion method for converting first moving image data encoded in accordance with a first motion compensated prediction method into second moving image data having the same format as data encoded in accordance with a second motion compensated prediction method. ,
A selection step of selecting a motion vector satisfying a condition to be a motion vector in the second motion compensation prediction method of the block from one or more motion vectors of the block included in the first moving image data;
An encoding step for performing encoding according to the second motion compensation prediction method, and an encoding step for encoding a block using each motion vector selected by the selecting step. Moving image conversion method. In a device capable of executing a computer program, the first moving image data encoded according to the first motion compensation prediction method is converted into second moving image data having the same format as the data encoded according to the second motion compensation prediction method. A computer program for performing video conversion processing for conversion,
A selection step of selecting a motion vector satisfying a condition to be a motion vector in the second motion compensation prediction method of the block from one or more motion vectors of the block included in the first moving image data;
An encoding step for performing encoding according to the second motion compensation prediction method, and an encoding step for encoding a block using each motion vector selected by the selecting step. Computer program. A computer-readable recording medium recording a computer program, wherein the computer program is
In a device capable of executing a computer program, the first moving image data encoded according to the first motion compensation prediction method is converted into second moving image data having the same format as the data encoded according to the second motion compensation prediction method. A computer program for performing video conversion processing for conversion,
A selection step of selecting a motion vector satisfying a condition to be a motion vector in the second motion compensation prediction method of the block from one or more motion vectors of the block included in the first moving image data;
An encoding step for performing encoding according to the second motion compensation prediction method, and an encoding step for encoding a block using each motion vector selected by the selecting step. To do. Integrated circuit of moving picture conversion apparatus for converting first moving picture data encoded according to first motion compensated prediction method into second moving picture data having the same format as data encoded according to second motion compensated prediction method Because
Selecting means for selecting a motion vector satisfying a condition to be a motion vector in the second motion compensated prediction method of the block from one or more motion vectors of the block included in the first moving image data;
Encoding means for performing encoding according to a second motion compensation prediction method, comprising: encoding means for encoding a block using each motion vector selected by the selection means; Integrated circuit.

Images