JP2006279272A - Moving picture coder and coding control method thereof - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a moving picture coder and a coding control method thereof for enhancing a coding efficiency by eliminating an isolated DCT coefficient. <P>SOLUTION: When the number of DCT (Discrete Cosine Transform ) coefficients below a dead zone wherein a quantization value is 0 in the moving picture coder is consecutive for N(f) or over, the dead zone is increased by an increment width (f) or over. On the other hand, when coefficients in excess of the dead zone (quantization value >0) appears, the dead zone is restored to the initial value. The consecutive number N(f) wherein the quantization value is 0 and the increment width T(f) of the dead zone are changed with the frequency f. The dead zone is introduced in this way and the DCT coefficients below the dead zone are eliminated for carrying out the quantization. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to a moving picture coding apparatus having means for performing discrete cosine transform and quantization and a coding control method thereof, and more specifically, a continuous number of frequency components equal to or smaller than a predetermined value (quantized value 0), and The present invention relates to a moving picture coding apparatus that derives a dead zone based on a frequency and a coding control method thereof.

  In recent years, H.264 has been introduced as a next-generation moving image encoding method. The H.264 standard is drawing attention.

  This H. In the H.264 standard, a considerable improvement (about twice or more) can be expected in terms of compression rate and image quality as compared with conventional systems such as MPEG (Moving Picture Experts Group) -2 and -4.

  However, moving picture coding schemes such as MPEG-2 adopt a technique called Q matrix that changes the quantization width of DCT (Discrete Cosine Transform) coefficients according to the frequency to improve the quality of the coded picture. H. H.264 has no corresponding technology.

Conventionally, what was described in the nonpatent literature 1 is known as what coped with such a problem. Non-Patent Document 1 mentions a technique for changing a quantization dead zone (a numerical group equal to or smaller than this value and having a DCT coefficient quantization value of 0) according to the frequency. While conforming to the H.264 standard, the effect of improving the subjective quality of the encoded image can be obtained in the same manner as the Q matrix used in MPEG-2 or the like.
Chono, Miyamoto, “Improvement of AVC image quality by quantization with frequency-dependent dead zone”, (2004 IEICE General Conference D-11-50)

  However, in the one described in Non-Patent Document 1, the quantization dead zone is fixed according to the frequency. For this reason, optimal quantization cannot be performed in accordance with the nature of the DCT coefficient to be encoded, leading to a decrease in encoding efficiency.

  FIG. 5 is a diagram showing how to adopt a dead zone according to the configuration of Non-Patent Document 1.

  According to this configuration, the quantization dead zone is fixedly determined according to the frequency. For this reason, if there is a wide-area DCT coefficient that exceeds this dead zone, it remains as an isolated DCT coefficient. Such an isolated DCT coefficient requires a large number of coding bits for coding, which causes a reduction in coding efficiency.

  Accordingly, the present invention has been made in view of the above problems, and a moving picture encoding apparatus and an encoding control method thereof that improve encoding efficiency by adaptively removing isolated DCT coefficients. The purpose is to provide.

  In order to achieve the above object, a first aspect of the present invention is a moving picture coding apparatus having means for performing discrete cosine transform and quantization, and performing discrete cosine transform on a residual signal predicted for moving picture data. Conversion means for converting to frequency components, a frequency component converted by the conversion means, a dead zone control means for deriving a dead zone according to the frequency corresponding to the frequency component, and a dead zone control means And a quantization means for performing quantization by removing frequency components converted by the conversion means below the dead zone.

  According to a fourth aspect of the present invention, there is provided a coding control method for a moving picture coding apparatus having means for performing discrete cosine transformation and quantization, and performing discrete cosine transformation on a residual signal predicted for moving picture data. Then, a dead zone is derived by the dead zone control unit according to the frequency component converted by the conversion unit and the frequency corresponding to the frequency component, and is derived by the dead zone control unit. The frequency component converted by the conversion unit below the dead zone is removed, and quantization is performed by the quantization unit.

  According to the present invention, moving image data is converted into a frequency component, a dead zone is derived according to the converted frequency component and a frequency corresponding to the frequency component, and a frequency equal to or less than the derived dead zone. Since the quantization is performed by removing the components, isolated wide-area DCT coefficients that cause a decrease in coding efficiency can be removed, so that the coding efficiency can be improved. Therefore, H.H. Even in the H.264 standard, an effect equivalent to the Q matrix can be obtained.

  Embodiments of a moving picture coding apparatus and a coding control method thereof according to the present invention will be described below in detail with reference to the accompanying drawings.

  FIG. 1 is a diagram showing an example of a schematic configuration of a moving picture encoding apparatus 10 according to the present invention. Note that this moving image encoding apparatus 10 is an H.264 format. The image data is encoded based on the H.264 / AVC standard.

  In FIG. 1, the image data supplied to the input terminal 101 is supplied to the subtracter 102. The subtracter 102 subtracts the image data from the switch 103 from the input image data when the inter-frame encoding process is being performed. The output data of the subtractor 102 is subjected to discrete cosine transform processing and quantization processing in a DCT (discrete cosine transform) and quantization unit 104. The output of the DCT / quantization unit 104 is variable-length encoded by an entropy encoding unit (may be referred to as a variable-length encoding unit) 105 and is output to the output terminal 106 as a stream.

  The output of the DCT / quantization unit 104 is input to the inverse quantization / inverse DCT unit 107 and inversely transformed. The inversely converted data is added to the image data from the switch 103 in the adder 108, reproduced as a frame image, and output. The output of the adder 108 is input to a deblocking filter 109 in order to improve distortion generated between blocks of image data blocked for DCT processing and quantization processing. The image data output from the deblocking filter 109 is input to the image memory 110.

  The motion compensation unit 111 reads the encoded image in the image memory 110 based on the image motion vector from the motion vector detection unit 112, and generates predicted image data. That is, the predicted image is generated from the motion information so that the already encoded image stored in the image memory 110 approaches the image of the input terminal 101.

  The motion vector detection unit 113 detects a motion vector indicating the motion of the moving image using the input image data of the input terminal 101. Since the motion vector is also referred to on the data side, it is sent to the entropy coding unit 105 as supplementary information and inserted into the header of a predetermined transmission unit.

  The output image data of the motion compensation unit 111 is given to the subtracter 102 via the switch 103. Since the output image data of the motion compensation unit 111 is predicted to be as close as possible to the input image data, the amount of data output from the subtractor 102 is efficiently reduced. In other words, this means that the compression efficiency is high.

  Here, when there is a periodic or scene change, an in-screen compression process is executed. At this time, the intra prediction unit 114 performs intra prediction from the already encoded pixels around the block to be encoded, and the subtracter 102 subtracts the intra prediction signal from the input image data at the input terminal 101. , DCT and quantization unit 104.

  In this way, image compression processing within one frame is executed in a loop formed by the DCT / quantization unit 104, the intra-screen prediction unit 114, the switch 103, and the subtractor 102. Image data (also referred to as an I (Intra) slice) compressed in a frame is inversely transformed and decoded by an inverse quantization and DCT unit 107, and this decoded data is deblocked by a deblocking filter 109. Is reduced and stored in the image memory 110. The image data at this time is compressed image data using only data within a frame, and serves as reference data when a moving image for a plurality of frames (in units of pictures) is reproduced.

  Next, a schematic configuration of the DCT and quantization unit 104 described in FIG. 1 will be described with reference to FIG.

  The DCT 11 plays a role of converting input image data into frequency components (DCT coefficients). Specifically, the input image data is converted into DCT coefficients using a discrete cosine function, and the DCT coefficients are output to the dead zone control unit 12 and the quantization unit 13.

  The dead zone control unit 12 derives a dead zone (a numerical group in which the quantized value of the frequency component equal to or less than this value is 0) based on the continuous number of DCT coefficients equal to or less than a predetermined value and the frequency. The term “below the predetermined value” refers to a DCT coefficient that is as close to 0 as possible, and a DCT coefficient that has a quantized value of 0.

  The quantization unit 13 quantizes the DCT coefficient input from the DCT 11 using a dead zone. Specifically, from the DCT coefficient input from the DCT 11, DCT coefficients equal to or less than the dead zone derived by the dead zone control unit 12 are removed (the quantization value is 0), while the DCT coefficients exceeding the dead zone (quantum) Quantization is performed on the quantization value> 0).

  Here, the outline of the dead zone-added quantization process in the dead zone control unit 12 and the quantization unit 13 will be described with reference to FIG.

  In the quantization process with a dead zone, the dead zone is increased and the width T (f) is increased when N (f) or more DCT coefficients close to 0 continue as long as the quantization value becomes zero. On the other hand, when a coefficient exceeding the dead zone (quantization value> 0) appears, the dead zone is returned to the initial value (default).

  As a result, it becomes possible to remove the isolated DCT coefficient (in this figure, the removed coefficient I1), and the coding efficiency is improved. Note that the continuous number N (f) of DCT coefficients (zero coefficients) with a quantized value 0 and the dead zone increase width T (f) are set according to the frequency.

  Next, the dead zone derivation process in the dead zone control unit 12 will be described.

  FIG. 4 is a flowchart for explaining the operation of the dead zone derivation process.

  When this process is started, an initial setting process is first performed. Specifically, the initial value 0 is set to the frequency: f and the number of zero coefficients: n, and the default value dzi (f) is set to the dead zone: dz (step S101).

  When the initial setting process is finished, it is determined whether or not the DCT coefficient: C (f) exceeds the dead zone: dz (step S102). If it is determined by this determination that the DCT coefficient: C (f) exceeds the dead zone: dz (YES in step S102), the zero coefficient number: 0 is set to n, and the default value is set to the dead zone: dz. : Dzi (f) is set (step S104), and the process proceeds to step S107. On the other hand, if it is determined that the DCT coefficient C (f) is equal to or less than the dead zone dz (NO in step S102) ), 1 is added to the number of zero coefficients: n (step S103), and then it is determined whether the number of zero coefficients: n exceeds the zero coefficient number threshold value: N (f) (step S105).

  If it is determined by this determination that the number of zero coefficients: n has exceeded the zero coefficient number threshold: N (f) (NO in step S105), the process proceeds to step S107. On the other hand, when it is determined that the zero coefficient number: n does not exceed the zero coefficient number threshold value: N (f) (YES in step S105), the dead zone: dz is increased by adding the width: T (f). (Step S106), the dead zone is raised.

  When the dead zone derivation process corresponding to the frequency f is thus completed, 1 is added to the frequency f (step S107), and the dead zone derivation process corresponding to the next frequency is performed.

  When 1 is added to the frequency: f, it is determined whether or not the frequency f after the addition exceeds the frequency end value F (step S108). That is, it is a determination as to whether or not the dead zone derivation process corresponding to all frequencies has been completed.

  If it is determined by this determination that the dead zone derivation processing corresponding to all frequencies has been completed (YES in step S108), this processing is terminated, but otherwise (NO in step S108). Returning to step S102, it is again determined whether or not the DCT coefficient C (f) corresponding to the next frequency exceeds the dead zone dz.

  As described above, in the present invention, by deriving a dead zone according to both the number of continuous DCT coefficients (zero coefficients) with a quantized value of 0 and the frequency, a factor that reduces coding efficiency is Since the isolated wide-area DCT coefficient can be removed, the encoding efficiency can be improved. Therefore, H.H. Even in the H.264 standard, an effect equivalent to the Q matrix can be obtained.

  The moving picture coding apparatus and the coding control method thereof according to the present invention are H.264 for moving picture data. The present invention can be applied to all moving picture coding apparatuses that perform coding processing conforming to the H.264 standard, and can be effectively used particularly when coding efficiency is improved by removing isolated coefficients.

It is a figure which shows an example of schematic structure of the moving image encoder 10 concerning this invention. FIG. 2 is a diagram illustrating a schematic configuration of a DCT and quantization unit 104 described in FIG. 1. It is a figure which shows the outline | summary of the quantization process with a dead zone in the dead zone control part 12 and the quantization part 13. FIG. 6 is a flowchart for explaining an operation of a dead zone derivation process in the dead zone control unit 12. It is a figure explaining the outline | summary of the quantization process with a conventional dead zone.

Explanation of symbols

10 video encoding device 11 DCT
DESCRIPTION OF SYMBOLS 12 Dead zone control part 13 Quantization part 101 Input image data 102 Subtractor 103 Switch 104 DCT and quantization part 105 Lower entropy code 106 Output terminal 107 Inverse DCT and inverse quantization part 108 Adder 109 Deblocking filter 110 Image memory 111 Motion compensation unit 112 Weighted prediction unit 113 Motion spectrum detection unit 114 Coding control unit

Claims (4)

In a moving image encoding apparatus having means for performing discrete cosine transform and quantization,
Transform means for transforming the residual signal predicted for moving image data into a frequency component by performing discrete cosine transform;
Dead zone control means for deriving a dead zone according to the frequency component converted by the conversion means and the frequency corresponding to the frequency component;
A moving picture coding apparatus comprising: quantization means for performing quantization by removing frequency components converted by the conversion means equal to or less than the dead zone derived by the dead zone control means. The dead zone control means includes
2. The moving picture coding apparatus according to claim 1, wherein, when a predetermined number or more of frequency components are continuous, a dead width is increased by an amount corresponding to a frequency corresponding to the frequency component. . The moving picture encoding apparatus according to claim 2, wherein when there is a frequency component exceeding the dead zone, the dead zone is returned to an initial value. In a coding control method of a moving picture coding apparatus having means for performing discrete cosine transform and quantization,
The residual signal predicted for moving image data is subjected to discrete cosine transform and converted to a frequency component by a conversion means,
A dead zone is derived by the dead zone control unit according to the frequency component converted by the conversion unit and the frequency corresponding to the frequency component,
A coding control method for a moving picture coding apparatus, characterized in that a frequency component transformed by the transforming means equal to or less than the dead zone derived by the dead zone control means is removed and quantization is performed by quantization means.

Images