JP2013158041A - Transcoder - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a technique that enables more minute image quality adjustment and code amount adjustment at a transcoder that decodes an input image into a decoded image and encodes the decoded image into an output image.SOLUTION: A quantization step value determination part 6 determines an output quantization step value for each macro block by correcting an input quantization step value based on various characteristic evaluation values. A CPU 5 controls the quantization step value determination part 6 in accordance with an input/output state based on an input bit rate, an output bit rate, and a target bit rate. The transcoder 1 is capable of performing image quality adjustment and code amount adjustment more minutely. As the various characteristic evaluation values, it is possible to cite an evaluation value indicative of a change degree in a space direction, an evaluation value indicative of a change degree in a time direction, an evaluation value indicative of a prediction error in the time direction, an evaluation value indicative of skin color chromaticity, and an evaluation value indicative of a coordinate position of each macro block.

Description

  The present invention relates to a transcoder that decodes an input image into a decoded image and encodes the decoded image into an output image.

  Image compression technology is widely used to reduce the transmission load and storage load of image data. MPEG2 or the like exists as a conventional encoding method, and H.264 is a new encoding method. H.264 exists. The transcoder performs code conversion between different or the same encoding methods in order to reduce the transmission load and storage load of image data.

  That is, the transcoder first inputs a compressed image by the first encoding method, and generates a decompressed image by the first encoding method. The transcoder then compresses the decompressed image by the second encoding method and outputs a compressed image by the second encoding method.

  Here, the transcoder needs to reduce the output bit rate to the target bit rate in order to reduce the transmission burden and storage burden of image data. However, the transcoder needs not to excessively increase the quantization step value in the second encoding method from the quantization step value in the first encoding method in order to reduce deterioration in image quality of the image data.

  The transcoder disclosed in Patent Document 1 reduces the image quality degradation of image data and reduces the transmission burden and storage burden of image data. That is, the transcoder calculates the degree of change in the space / time direction for each macroblock. Then, for each macroblock, the transcoder corrects the quantization step value in the first coding scheme based on the degree of change in the space / time direction, and determines the quantization step value in the second coding scheme. .

  That is, the transcoder allocates a large amount of code of a compressed image by the second encoding method for an image area where the image quality of the image data needs to be increased. Then, the transcoder does not allocate a large amount of code of the compressed image by the second encoding method for an image area where it is not necessary to improve the image quality of the image data. Therefore, the transcoder can reduce the code amount of the entire compressed image by the second encoding method.

JP 2008-42426 A

  The transcoder disclosed in Patent Document 1 reduces the image quality degradation of image data and reduces the transmission burden and storage burden of image data. That is, for each macroblock, the transcoder corrects the quantization step value in the first coding scheme based on the degree of change in the space / time direction, and determines the quantization step value in the second coding scheme. . However, it is expected that the transcoder can finely adjust the image quality and the code amount by taking into account the evaluation based on other parameters.

  Therefore, in view of the above problems, the present invention provides a technique capable of finely adjusting image quality adjustment and code amount adjustment in a transcoder that decodes an input image into a decoded image and encodes the decoded image into an output image. For the purpose.

In order to solve the above-mentioned problem, the invention 1 is a transcoder comprising: a decoder that decodes an input image to generate a decoded image; and an encoder that encodes the decoded image to generate an output image. A spatial change evaluation value calculation unit that calculates a spatial change evaluation value that indicates the degree of change in the spatial direction in a prediction error evaluation value calculation unit that calculates a prediction error evaluation value that indicates a prediction error in the time direction in each macroblock, Quantization that corrects the input quantization step value of each macroblock and determines the output quantization step value of each macroblock by using a correction coefficient based on the spatial change evaluation value and the prediction error evaluation value A step value determining unit, wherein the correction coefficient is multiplied by an input quantization step value to determine an output quantization step value. And coefficient, in order to output the quantization step value is determined, characterized in that it comprises a, and adding the correction factor to be added to the input quantization step value.

A second aspect of the invention is the transcoder of the first aspect, further comprising: a quantization step value obtained by multiplying an input quantization step value by the multiplication correction coefficient; and a quantization step in which the addition correction coefficient is added to an input quantization step value. Any quantization step of a value and a quantization step value in which the multiplication correction coefficient is not multiplied by an input quantization step value and the addition correction coefficient is not added to an input quantization step value An output quantization step value selection unit that selects a value as an output quantization step value is provided.

A third aspect of the present invention is transcoder Invention 1 or Invention 2, the quantization step value determining unit, the smaller the spatial change evaluation value, to determine low output quantization step value, the larger the prediction error evaluation value As a result, the output quantization step value is determined to be small.

A fourth aspect of the present invention is any of the transcoder of the invention 1 to invention 3, the quantization step value determining unit, based on the picture type of each macroblock, and determines the output quantization step value.

A fifth aspect of the present invention is any of the transcoder of the invention 1 to invention 4, the spatial variation evaluation value in each macroblock of the decoded image, a difference absolute value of the average pixel value and the pixel value, obtained by adding for each pixel Intra-image difference absolute value sum is included.

Invention 6, in any of the transcoder of the invention 1 to invention 5, the prediction error evaluation value in each macroblock of the decoded image and the reference image, the difference absolute value of each corresponding pixel value, added up for each corresponding pixel The sum of absolute differences between images is included.

Invention 7, the invention in one to one of transcoders invention 6, the quantization step value determining unit, the ratio of the prediction error evaluation value and the spatial variation evaluation value as a variable, the correction for storing said correction factor A coefficient storage unit.

An invention 8 is the transcoder according to any one of the invention 1 to the invention 6, wherein the quantization step value determination unit stores the correction coefficient using the spatial change evaluation value and the prediction error evaluation value as variables. Part.

The invention 9 is a transcoder comprising a decoder that decodes an input image to generate a decoded image, and an encoder that encodes the decoded image to generate an output image, and the degree of change in the spatial direction in each macroblock is determined. A spatial change evaluation value calculation unit that calculates a spatial change evaluation value, a time change evaluation value calculation unit that calculates a time change evaluation value that indicates the degree of change in the time direction in each macroblock, the spatial change evaluation value, and the time A quantization step value determination unit that corrects an input quantization step value of each macroblock by using a correction coefficient based on the change evaluation value and determines an output quantization step value of each macroblock; It is characterized by providing.

A tenth aspect of the invention is the transcoder of the ninth aspect, wherein the quantization step value determining unit determines an output quantization step value to be smaller as the spatial change evaluation value is larger, and an output is larger as the time change evaluation value is larger. The quantization step value is determined to be small.

An eleventh aspect of the invention is characterized in that in the transcoder of the ninth or tenth aspect of the invention , the quantization step value determining unit determines an output quantization step value based on a picture type of each macroblock.

Invention 12, in any one of the transcoder of the invention 9 to invention 11, the spatial variation evaluation value in each macroblock of the decoded image, a difference absolute value of the average pixel value and the pixel value, obtained by adding for each pixel Intra-image difference absolute value sum is included.

Invention 13, in any one of the transcoder of the invention 9 to invention 12, wherein the time variation evaluation value, characterized in that it comprises a size of the motion vector in each macroblock of the input image.

Invention 14, in any one of the transcoder of the invention 9 to invention 12, wherein the time variation evaluation value, characterized in that it comprises a code amount of motion vectors in each macroblock of the input image.

Invention 15, the invention 9 to in any of the transcoders invention 14, the quantization step value determining unit, the time variation evaluation value and the spatial variation evaluation value as a variable, stored correction coefficients for storing said correction factor Part.

A sixteenth aspect of the present invention is a transcoder comprising a decoder that decodes an input image to generate a decoded image, and an encoder that encodes the decoded image to generate an output image, wherein the degree of change in the spatial direction in each macroblock is determined. A spatial change evaluation value calculating unit that calculates a spatial change evaluation value to be indicated; a predetermined chromaticity evaluation value calculating unit that calculates a predetermined chromaticity evaluation value indicating a predetermined chromaticity in each macroblock; the spatial change evaluation value and the predetermined Based on the chromaticity evaluation value, by using a correction coefficient, a quantization step value determination unit that corrects an input quantization step value of each macroblock and determines an output quantization step value of each macroblock; It is characterized by providing.

A seventeenth aspect of the present invention is the transcoder of the sixteenth aspect, wherein the predetermined chromaticity evaluation value includes a flesh color chromaticity evaluation value indicating a flesh color chromaticity in each macroblock of a decoded image, and the quantization step value determining unit includes the The smaller the spatial change evaluation value, the smaller the output quantization step value, and the smaller the skin color chromaticity evaluation value, the smaller the output quantization step value.

The invention 18 is the transcoder of the invention 16 or the invention 17, wherein the spatial change evaluation value is an intra-image difference obtained by adding a difference absolute value between each pixel value and an average pixel value for each pixel in each macroblock of the decoded image. A sum of absolute values.

A nineteenth aspect of the invention is the transcoder according to any one of the sixteenth to eighteenth aspects of the invention , wherein the quantization step value determining unit stores the correction coefficient using the spatial change evaluation value and the predetermined chromaticity evaluation value as variables. A storage unit.

Invention 20 is a transcoder comprising a decoder that decodes an input image to generate a decoded image, and an encoder that encodes the decoded image to generate an output image, wherein the coordinate position evaluation indicates the position of each macroblock. A coordinate position evaluation value calculation unit for calculating a value, and using the correction coefficient based on the coordinate position evaluation value, the input quantization step value of each macroblock is corrected, and the output quantization of each macroblock A quantization step value determination unit that determines a step value.

Invention 21, in the transcoder of the invention 20, the quantization step value determining unit, the position of each macro block is closer to the position of the image center, and determines reduce the output quantization step value.

Invention 22, in the transcoder of the invention 20 or invention 21, the quantization step value determining unit, the coordinate position evaluation value as a variable, characterized in that it comprises a correction coefficient storage unit, for storing said correction factor .

Invention 23, the invention 7, the invention 8, the invention 15, the invention 19, in any of the transcoders invention 22 further comprises an input bit rate of the input image, the output bit rate of the output image, the target bit rate of the output image And a correction coefficient update writing unit that updates and writes the correction coefficient to the correction coefficient storage unit every moment based on the time change.

Invention 24, the invention in any of the transcoder 1 to invention 23, further the evaluation value for each macroblock, comprising an evaluation value division divided unit, to perform division classification of a predetermined number, the quantization step The value determining unit determines the output quantization step value based on the evaluation value that has been subjected to the predetermined number of divisions.

A twenty-fifth aspect of the present invention is the transcoder of the twenty-fourth aspect, wherein the ranges of evaluation values in each section are equal.

The invention 26 is characterized in that, in the transcoder of the invention 24, the range of the evaluation value in each section is different.

Invention 27, in any one of the transcoder of the invention 9 to invention 22, the correction coefficient, in order to output the quantization step value is determined, the multiplication correction factor to be multiplied to an input quantization step value, the output quantum In order to determine the quantization step value, an addition correction coefficient added to the input quantization step value is included.

A twenty-eighth aspect of the present invention is the transcoder of the twenty-seventh aspect, further comprising: a quantization step value obtained by multiplying an input quantization step value by the multiplication correction coefficient; and a quantization step in which the addition correction coefficient is added to an input quantization step value. Any quantization step of a value and a quantization step value in which the multiplication correction coefficient is not multiplied by an input quantization step value and the addition correction coefficient is not added to an input quantization step value An output quantization step value selection unit that selects a value as an output quantization step value is provided.

The invention 29 is the transcoder according to any one of the invention 1 to the invention 28, wherein the quantization step value determining unit is finely adjusted by a fine adjustment coefficient with respect to the output quantization step value corrected by the correction coefficient. A fine adjustment unit for calculating an output quantization step value.

A thirty-third aspect of the invention is characterized in that, in the transcoder of the twenty-ninth aspect, the fine adjustment coefficient includes a multiplication fine adjustment coefficient that is multiplied by an output quantization step value corrected by the correction coefficient.

Invention 31, in the transcoder of the invention 29 or invention 30, the fine adjustment factor, characterized in that it comprises a summing fine adjustment factor, which is added to the correction output quantization step value by the correction coefficient.

Invention 32, in any one of the transcoder of the invention 29 to invention 31, the fine adjustment factor based on the picture type of each macroblock, characterized in that it is set respectively.

A thirty-third aspect of the invention is the transcoder according to any one of the twenty-ninth to thirty-second aspects of the present invention , wherein the fine adjustment coefficient is a time change of an input bit rate of the input image, an output bit rate of the output image, and a target bit rate of the output image Based on the above, it is updated every moment.

  The transcoder of the present invention determines an output quantization step value by correcting an input quantization step value based on various characteristic evaluation values for each macroblock. Thereby, the transcoder can finely adjust the image quality and the code amount. Here, the following characteristic evaluation values are given as various characteristic evaluation values.

  Examples of the first characteristic evaluation value include a spatial change evaluation value indicating the degree of change in the spatial direction and a prediction error evaluation value indicating a prediction error in the time direction. The smaller the spatial change evaluation value and the larger the prediction error evaluation value, the smaller the output quantization step value is determined. As a result, the image quality can be further improved for an image region having a large prediction error in the time direction. Even if the prediction error in the time direction is small, the image quality can be further improved for an image region having a small degree of change in the spatial direction. The code amount of the entire image is also adjusted.

  Examples of the second characteristic evaluation value include a space change evaluation value indicating the degree of change in the spatial direction and a time change evaluation value indicating the degree of change in the time direction. The larger the space change evaluation value and the larger the time change evaluation value, the smaller the output quantization step value is determined. As a result, the image quality can be further improved for an image region having a large degree of change in the spatial direction and a large degree of change in the time direction. The code amount of the entire image is also adjusted.

  Examples of the third characteristic evaluation value include a spatial change evaluation value indicating the degree of change in the spatial direction and a skin color chromaticity evaluation value indicating the skin color chromaticity. The smaller the space change evaluation value is, and the larger the skin color chromaticity evaluation value is, the smaller the output quantization step value is determined. As a result, the image quality of the face image area can be further improved. The code amount of the entire image is also adjusted.

  As the fourth characteristic evaluation value, there is a coordinate position evaluation value indicating the coordinate position of each macroblock. The closer the coordinate position of each macroblock is to the coordinate position of the center of the image, the smaller the output quantization step value is determined. As a result, the image quality of the central image region can be further improved. The code amount of the entire image is also adjusted.

It is a block diagram which shows the component of a transcoder. It is a block diagram which shows the component of a quantization step value determination part. It is a block diagram which shows the component of a 1st determination part. It is a figure which shows the content of LUT in a 1st determination part. It is a block diagram which shows the component of a 1st determination part. It is a figure which shows the categorization method of an evaluation value. It is a figure which shows the content of LUT in a 1st determination part. It is a block diagram which shows the component of a 2nd determination part. It is a figure which shows the content of LUT in a 2nd determination part. It is a flowchart which shows the flow of a process in a logic circuit part. It is a block diagram which shows the component of a 3rd determination part. It is a figure which shows the content of LUT in a 3rd determination part. It is a block diagram which shows the component of a 4th determination part. It is a figure which shows the content of LUT in a 4th determination part. It is a block diagram which shows the component of a quantization step value determination part. It is a block diagram which shows the component of a 5th determination part. It is a block diagram which shows the component of a quantization step value determination part. It is a block diagram which shows the component of a fine adjustment part.

{First embodiment}
[Transcoder components]
Hereinafter, the first embodiment will be described with reference to the drawings. FIG. 1 is a block diagram showing components of the transcoder 1. The transcoder 1 inputs an input image encoded by MPEG2, An output image encoded by H.264 is output. The encoding method of the input image and the output image is MPEG2 or H.264. It is not limited to H.264. The encoding method of the input image and the output image may be different or the same.

  The transcoder 1 includes a decoder 2, an encoder 3, a process setting unit 4, a CPU 5, a quantization step value determining unit 6, an evaluation value calculating unit 7, an image memory 8, and the like.

  The decoder 2 decodes the input image and generates a decoded image. That is, the decoder 2 performs decoding, inverse quantization, and frequency inverse transform in the above order. The encoder 3 encodes the decoded image and generates an output image. That is, the encoder 3 performs frequency conversion, quantization, and encoding in the above order.

  The process setting unit 4 receives process settings such as a target bit rate from the user of the transcoder 1. The CPU 5 comprehensively controls the transcoder 1 based on the input bit rate notified from the decoder 2, the output bit rate notified from the encoder 3, and the target bit rate notified from the processing setting unit 4.

  The quantization step value determination unit 6 inputs and corrects the quantization step value of the input image from the decoder 2 for each macroblock. The quantization step value determination unit 6 determines the quantization step value of the output image for each macroblock and outputs it to the encoder 3.

  The evaluation value calculation unit 7 inputs a decoded image from the decoder 2 and calculates various characteristic evaluation values (described later) for each macroblock. The image memory 8 temporarily stores the decoded image when the decoded image is input from the decoder 2 and output to the encoder 3.

  The correction of the quantization step value of the input image and the determination of the quantization step value of the output image are notified from the correction coefficient (described later) stored in the quantization step value determination unit 6 and the evaluation value calculation unit 7. It is executed based on various characteristic evaluation values (described later).

[Components of quantization step value determination unit]
FIG. 2 is a block diagram illustrating components of the quantization step value determination unit 6. FIG. 2 includes peripheral components of the quantization step value determination unit 6. The quantization step value determination unit 6 includes a first determination unit 61, a second determination unit 62, a third determination unit 63, a fourth determination unit 64, and the like.

  The first determination unit 61 inputs the quantization step value of the input image input to the transcoder 1 from the decoder 2 for each macroblock, and outputs the corrected quantization step value to the second determination unit 62. . The first determination unit 61 corrects the quantization step value of the input image input to the transcoder 1 for each macroblock, the spatial change evaluation value indicating the degree of change in the spatial direction, and the prediction error in the temporal direction. A prediction error evaluation value indicating is used.

  The second determination unit 62 inputs the quantization step value corrected by the first determination unit 61 for each macroblock from the first determination unit 61 and supplies the corrected quantization step value to the third determination unit 63. Output. When correcting the quantization step value corrected by the first determining unit 61 for each macroblock, the second determining unit 62 calculates the spatial change evaluation value indicating the change degree in the spatial direction and the change degree in the time direction. The time change evaluation value shown is used.

  For each macroblock, the third determination unit 63 inputs the quantization step value corrected by the second determination unit 62 from the second determination unit 62, and sends the corrected quantization step value to the fourth determination unit 64. Output. The third determining unit 63 corrects the quantization step value corrected by the second determining unit 62 for each macroblock, and the skin change evaluation value indicating the degree of change in the spatial direction and the skin color indicating the skin color chromaticity Use the chromaticity evaluation value.

  The fourth determination unit 64 inputs the quantization step value corrected by the third determination unit 63 for each macroblock from the third determination unit 63 and outputs the quantization step value of the output image output from the transcoder 1. Is output to the encoder 3. The fourth determination unit 64 uses a coordinate position evaluation value indicating the coordinate position of each macroblock when correcting the quantization step value corrected by the third determination unit 63 for each macroblock.

  In the following description of the first embodiment, the processing contents of the first determination unit 61, the second determination unit 62, the third determination unit 63, and the fourth determination unit 64 will be described in the above order.

[Processing content of first determination unit 61A]
FIG. 3 is a block diagram showing components of the first determination unit 61A as a kind of the first determination unit 61. As shown in FIG. The first determination unit 61A inputs “input quantization step value”, “ACT_MB”, and “SAD_MB” at the left end of FIG. The first determination unit 61A outputs an “output quantization step value” at the right end of FIG.

  The “input quantization step value” is a quantization step value of an input image input to the transcoder 1 in each macroblock. The “output quantization step value” is a quantization step value corrected by the first determination unit 61A in each macroblock.

  “ACT_MB” is a spatial change evaluation value indicating the degree of change in the spatial direction in each macroblock. The space change evaluation value ACT_MB is calculated as described below. First, the evaluation value calculation unit 7 inputs a decoded image from the decoder 2. Next, the evaluation value calculation unit 7 calculates the spatial change evaluation value ACT_MB by adding the absolute value of the difference between each pixel value and the average pixel value for each pixel in each macroblock of the decoded image.

  “SAD_MB” is a prediction error evaluation value indicating a prediction error in the time direction in each macroblock. The prediction error evaluation value SAD_MB is calculated as described below. First, the decoder 2 searches for a reference image referred to by the decoded image. Next, the decoder 2 calculates the prediction error evaluation value SAD_MB by adding the difference absolute value of each corresponding pixel value for each corresponding pixel in each macroblock of the decoded image and the reference image.

  “P_type_MB” is a parameter indicating the picture type in each macroblock. Examples of picture types include intra macroblocks in I pictures, P pictures, B pictures, P pictures, or B pictures.

  When the picture type is an I picture, there is no reference picture referenced by that picture. At this time, the prediction error evaluation value SAD_MB in each corresponding macroblock of the P picture located immediately before the picture is adopted as the prediction error evaluation value SAD_MB in each macroblock of the picture.

  When the picture type is an intra macroblock in a P picture or a B picture, there is no reference macroblock referenced by the macroblock. At this time, the prediction error evaluation value SAD_MB in the macroblock located in the vicinity of the macroblock is adopted as the prediction error evaluation value SAD_MB in the macroblock.

  The first determination unit 61A includes shifters 611A and 612A, a selector 613A, an adder 614A, a divider 615A, a correction coefficient storage unit 616A, a multiplier 617A, an adder 618A, a selector 619A, and the like. In particular, the correction coefficient storage unit 616A will be described.

  The correction coefficient storage unit 616A stores the correction coefficient in order to correct the input quantization step value and determine the output quantization step value. A multiplication coefficient as a correction coefficient is stored in the multiplication coefficient one-dimensional LUT 616AM with the ratio of the spatial change evaluation value ACT_MB to the prediction error evaluation value SAD_MB as one variable. An addition coefficient as a correction coefficient is stored in the addition coefficient one-dimensional LUT 616AA with the ratio of the spatial change evaluation value ACT_MB to the prediction error evaluation value SAD_MB as one variable.

  Here, strictly speaking, the multiplication coefficient and the addition coefficient are not stored with the ratio of the spatial change evaluation value ACT_MB to the prediction error evaluation value SAD_MB as one variable. Rather, strictly speaking, the ratio of the spatial change evaluation value ACT_MB shifted by the shifter 611A to the prediction error evaluation value SAD_MB shifted by the shifter 612A and added by the adder 614A is defined as one variable. Stores the addition coefficient. The reason why the multiplication coefficient and the addition coefficient are stored as described above will be described.

  In the first place, the distance between pictures of each macroblock and its reference macroblock differs depending on the picture type of each macroblock. Therefore, the prediction error evaluation value SAD_MB in each macroblock has a different range depending on the picture type of each macroblock. However, the multiplication coefficient one-dimensional LUT 616AM and the addition coefficient one-dimensional LUT 616AA do not depend on the picture type of each macroblock, and only one type is stored.

  Therefore, the prediction error evaluation value SAD_MB in each macroblock is subjected to a different shift operation depending on the picture type of each macroblock. Therefore, the prediction error evaluation value SAD_MB in each macroblock for which the shift operation has been performed does not differ in range depending on the picture type of each macroblock. As a result, even if only one type of multiplication coefficient one-dimensional LUT 616AM and addition coefficient one-dimensional LUT 616AA are stored, they are effectively used regardless of the picture type of each macroblock.

  Next, description will be made regarding the processing contents in which the input quantization step value is corrected and the output quantization step value is determined for a certain macroblock in the first determination unit 61A.

  The shifter 611A inputs the space change evaluation value ACT_MB for a certain macroblock. The shifter 611A performs a shift operation on the space change evaluation value ACT_MB by the number of shift bits act_shift_bit. By the shift operation by the shifter 611A, the multiplication coefficient one-dimensional LUT 616AM and the addition coefficient one-dimensional LUT 616AA are used more effectively without depending on the picture type of each macroblock.

  The shifter 612A inputs the prediction error evaluation value SAD_MB for a certain macroblock. The shifter 612A performs a shift operation on the prediction error evaluation value SAD_MB by the number of shift bits sad_shift_bit. Here, the shift bit number sad_shift_bit is determined as described below.

  The selector 613A inputs the picture type P_type_MB as a switching signal for a certain macroblock. When the picture type P_type_MB is an I picture, a P picture, or a B picture, the selector 613A outputs sad_shift_bit_I, sad_shift_bit_P, and sad_shift_bit_B as the number of shift bits, respectively. When the picture type P_type_MB is an intra macroblock of a P picture or a B picture, the selector 613A outputs sad_shift_bit_I_PB as the number of shift bits.

  The adder 614A adds the offset sad_offset to the prediction error evaluation value SAD_MB for which the shift operation has been performed. The prediction error evaluation value SAD_MB subjected to the shift operation is finely finely adjusted by the addition operation by the adder 614A.

  The divider 615A divides the spatial change evaluation value ACT_MB for which the shift operation has been performed by the prediction error evaluation value SAD_MB for which the shift operation and the addition operation have been performed, thereby obtaining the spatial change evaluation value ACT_MB for the prediction error evaluation value SAD_MB. Calculate the percentage.

  The correction coefficient storage unit 616A inputs a ratio of the spatial change evaluation value ACT_MB to the prediction error evaluation value SAD_MB. The correction coefficient storage unit 616A calculates a multiplication coefficient using the multiplication coefficient one-dimensional LUT 616AM. The correction coefficient storage unit 616A calculates an addition coefficient using the addition coefficient one-dimensional LUT 616AA.

  FIG. 4A shows the contents of the multiplication coefficient one-dimensional LUT 616AM. As the ratio of the spatial change evaluation value ACT_MB to the prediction error evaluation value SAD_MB increases, the multiplication coefficient increases monotonously. When the ratio of the spatial change evaluation value ACT_MB to the prediction error evaluation value SAD_MB is the predetermined value x, the multiplication coefficient is 1. That is, the quantization step value determined using the multiplication coefficient does not change from the input quantization step value.

  FIG. 4B shows the contents of the addition coefficient one-dimensional LUT 616AA. As the ratio of the spatial change evaluation value ACT_MB to the prediction error evaluation value SAD_MB increases, the addition coefficient increases monotonously. When the ratio of the spatial change evaluation value ACT_MB to the prediction error evaluation value SAD_MB is the predetermined value x, the addition coefficient is 0. That is, the quantization step value determined using the addition coefficient does not change from the input quantization step value.

  The multiplier 617A multiplies the input quantization step value by the multiplication coefficient calculated by the multiplication coefficient one-dimensional LUT 616AM. The adder 618A adds the addition coefficient calculated by the addition coefficient one-dimensional LUT 616AA to the input quantization step value.

  The selector 619A inputs any one of “0”, “1”, and “2” as a switching signal. When the switching signal is “0”, the selector 619A outputs the uncorrected input quantization step value as the output quantization step value. When the switching signal is “1”, the selector 619A outputs the input quantization step value multiplied by the multiplication coefficient as the output quantization step value. When the switching signal is “2”, the selector 619A outputs the input quantization step value added with the addition coefficient as the output quantization step value.

  Of the image regions having the same degree of change in the spatial direction, the output quantization step value is determined to be small for an image region having a large prediction error in the temporal direction. The output quantization step value is determined to be large for an image region with a small prediction error in the time direction.

  That is, for an image region having a large prediction error in the time direction, the image quality can be further improved by assigning a large number of output bits. For an image region with a small prediction error in the time direction, even if a large number of output bits are not assigned, the image quality is not excessively degraded. Thereby, the code amount adjustment of the whole image is also achieved.

  Of the image regions having the same prediction error in the temporal direction, the output quantization step value is determined to be small for an image region having a small degree of change in the spatial direction. For an image region having a large degree of change in the spatial direction, the output quantization step value is determined to be large.

  That is, for an image region having a small degree of change in the spatial direction, the image quality can be further improved by assigning a large number of output bits. For an image region having a large degree of change in the spatial direction, even if a large number of output bits are not assigned, the image quality is not excessively deteriorated. Thereby, the code amount adjustment of the whole image is also achieved.

  Here, the contents of the multiplication coefficient one-dimensional LUT 616AM and the addition coefficient one-dimensional LUT 616AA may be fixed without changing over time. Further, the number of shift bits act_shift_bit, sad_shift_bit, and offset sad_offset may be fixed without changing over time. Further, the switching signal to the selector 619A may be fixed without changing over time.

  However, the contents of the multiplication coefficient one-dimensional LUT 616AM and the addition coefficient one-dimensional LUT 616AA may be rewritten from time to time by the CPU 5 based on the input bit rate, the output bit rate, and the target bit rate. For example, when the output bit rate is higher than the target bit rate, the multiplication coefficient and the addition coefficient may be greatly rewritten.

  Further, the shift bit numbers act_shift_bit, sad_shift_bit, and offset sad_offset may be updated by the CPU 5 from time to time based on the input bit rate, the output bit rate, and the target bit rate. For example, when the output bit rate is higher than the target bit rate, the shift bit numbers act_shift_bit and sad_shift_bit may be updated so that the ratio of the spatial change evaluation value ACT_MB to the prediction error evaluation value SAD_MB increases.

  Further, the switching signal to the selector 619A may be updated every moment by the CPU 5 based on the input bit rate, the output bit rate, and the target bit rate. For example, when the output bit rate is higher than the target bit rate, the switching signal to the selector 619A is updated so that the maximum quantization step value among the quantization step values input to the selector 619A is output from the selector 619A. It is good to be done.

[Processing content of first determination unit 61B]
FIG. 5 is a block diagram showing components of the first determination unit 61B as a kind of the first determination unit 61. The first determination unit 61B inputs “input quantization step value”, “h_ACT_MB”, and “h_SAD_MB” at the left end of FIG. The first determination unit 61B outputs an “output quantization step value” at the right end of FIG.

  The “input quantization step value” is a quantization step value of an input image input to the transcoder 1 in each macroblock. The “output quantization step value” is a quantization step value corrected by the first determination unit 61B in each macroblock.

  “H_ACT_MB” is a spatial change evaluation value indicating the degree of change in the spatial direction in each macroblock. The space change evaluation value h_ACT_MB is a category number when the space change evaluation value ACT_MB is categorized. Here, the spatial change evaluation value ACT_MB is subjected to a shift operation similar to that performed by the shifter 611A in the evaluation value calculation unit 7.

  “H_SAD_MB” is a prediction error evaluation value indicating a prediction error in the time direction in each macroblock. The prediction error evaluation value h_SAD_MB is a category number when the prediction error evaluation value SAD_MB is categorized. Here, the prediction error evaluation value SAD_MB is subjected to a shift operation similar to the shifter 612A and an addition operation similar to the adder 614A in the evaluation value calculation unit 7.

  FIG. 6 is a diagram illustrating a method for categorizing evaluation values. As common points of FIGS. 6A, 6B, and 6C, the evaluation values are categorized into five categories from the minimum value to the maximum value. As the category number increases, the evaluation value belonging to the category increases.

  In the first categorization method shown in FIG. 6A, the ranges of categories 0, 1, 2, 3, and 4 are the same. At this time, categorization of evaluation values is easily performed.

  In the second categorization method shown in FIG. 6B, the ranges of categories 0, 1, 2, 3, and 4 are different. The category range increases as the category number increases. At this time, the evaluation value is categorized more finely as the evaluation value is closer to the minimum value.

  In the third categorization method shown in FIG. 6C, the categories 0, 1, 2, 3, and 4 have different ranges. As the category number approaches 2, the category range decreases. At this time, the evaluation value is categorized more finely as the evaluation value is closer to the intermediate value.

  The first determination unit 61B includes a correction coefficient storage unit 611B, a multiplier 612B, an adder 613B, a selector 614B, and the like. In particular, the correction coefficient storage unit 611B will be described.

  The correction coefficient storage unit 611B stores the correction coefficient in order to correct the input quantization step value and determine the output quantization step value. A multiplication coefficient as a correction coefficient is stored in the multiplication coefficient two-dimensional LUT 611BM with the prediction error evaluation value h_SAD_MB and the spatial change evaluation value h_ACT_MB as two variables. An addition coefficient as a correction coefficient is stored in the addition coefficient two-dimensional LUT 611BA with the prediction error evaluation value h_SAD_MB and the spatial change evaluation value h_ACT_MB as two variables. Since the correction coefficient storage unit 611B stores a two-dimensional LUT having two variables for the category number, the storage capacity can be reduced even if the two-dimensional LUT is stored.

  Next, description will be made regarding the processing contents in which the first quantization unit 61B corrects the input quantization step value and determines the output quantization step value for a certain macroblock.

  The correction coefficient storage unit 611B inputs a prediction error evaluation value h_SAD_MB and a spatial change evaluation value h_ACT_MB for a certain macroblock. The correction coefficient storage unit 611B calculates a multiplication coefficient using the multiplication coefficient two-dimensional LUT 611BM. The correction coefficient storage unit 611B calculates an addition coefficient using the addition coefficient two-dimensional LUT 611BA.

  FIG. 7A shows the contents of the multiplication coefficient two-dimensional LUT 611BM. As the prediction error evaluation value h_SAD_MB increases, the multiplication coefficient decreases monotonously. As the space change evaluation value h_ACT_MB increases, the multiplication coefficient increases monotonously.

  FIG. 7B shows the contents of the addition coefficient two-dimensional LUT 611BA. As the prediction error evaluation value h_SAD_MB increases, the addition coefficient decreases monotonously. As the space change evaluation value h_ACT_MB increases, the addition coefficient increases monotonously.

  The multiplier 612B multiplies the input quantization step value by the multiplication coefficient calculated by the multiplication coefficient two-dimensional LUT 611BM. The adder 613B adds the addition coefficient calculated by the addition coefficient two-dimensional LUT 611BA to the input quantization step value.

  The selector 614B inputs any one of “0”, “1”, and “2” as a switching signal. When the switching signal is “0”, the selector 614B outputs an uncorrected input quantization step value as the output quantization step value. When the switching signal is “1”, the selector 614B outputs the input quantization step value multiplied by the multiplication coefficient as the output quantization step value. When the switching signal is “2”, the selector 614B outputs the input quantization step value added with the addition coefficient as the output quantization step value.

  The contents of the multiplication coefficient two-dimensional LUT 611BM and the addition coefficient two-dimensional LUT 611BA may be fixed without changing over time, or may be rewritten by the CPU 5 every moment. The switching signal to the selector 619A may be fixed without changing with time, and may be updated by the CPU 5 every moment.

[Processing content of second determination unit]
FIG. 8 is a block diagram showing components of the second determination unit 62. The second determination unit 62 inputs “input quantization step value”, “h_ACT_MB”, “h_MVbit_MB”, and “P_type_MB” at the left end of FIG. The second determination unit 62 outputs an “output quantization step value” at the right end of FIG.

  The “input quantization step value” is a quantization step value corrected by the first determination unit 61 in each macroblock. The “output quantization step value” is a quantization step value corrected by the second determination unit 62 in each macroblock.

  “H_ACT_MB” is a spatial change evaluation value indicating the degree of change in the spatial direction in each macroblock. The space change evaluation value h_ACT_MB is calculated as described below. First, the evaluation value calculation unit 7 inputs a decoded image from the decoder 2. Next, the evaluation value calculation unit 7 calculates the spatial change evaluation value ACT_MB by adding the absolute value of the difference between each pixel value and the average pixel value for each pixel in each macroblock of the decoded image. Next, the evaluation value calculation unit 7 performs categorization by the categorization method shown in FIG.

  “H_MVbit_MB” is a time change evaluation value indicating the degree of change in the time direction in each macroblock. The time change evaluation value h_MVbit_MB is calculated as described below. First, the decoder 2 acquires the magnitude of the motion vector or the code amount of the motion vector in each macroblock of the input image. Here, the code amount of the motion vector corresponds to the magnitude of the motion vector. Next, the evaluation value calculation unit 7 performs categorization by the categorization method shown in FIG.

  “P_type_MB” is a parameter indicating the picture type in each macroblock. Examples of picture types include intra macroblocks in I pictures, P pictures, B pictures, P pictures, or B pictures.

  When the picture type is an I picture, there is no reference picture referenced by that picture. At this time, the temporal change evaluation value h_MVbit_MB in each macroblock of the picture is not calculated.

  When the picture type is an intra macroblock in a P picture or a B picture, there is no reference macroblock referenced by the macroblock. At this time, the time change evaluation value h_MVbit_MB in the macroblock is not calculated.

  The second determination unit 62 includes correction coefficient storage units 621, 622, and 623, a selector 624, a logic circuit unit 625, a multiplier 626, an adder 627, a selector 628, and the like. In particular, the correction coefficient storage units 621, 622, and 623 will be described.

  The correction coefficient storage unit 621 stores a correction coefficient for correcting an input quantization step value and determining an output quantization step value for a P picture. A multiplication coefficient as a correction coefficient is stored in the multiplication coefficient two-dimensional LUT 621M with the space change evaluation value h_ACT_MB and the time change evaluation value h_MVbit_MB_P as two variables. An addition coefficient as a correction coefficient is stored in the addition coefficient two-dimensional LUT 621A with the space change evaluation value h_ACT_MB and the time change evaluation value h_MVbit_MB_P as two variables.

  The correction coefficient storage unit 622 stores the correction coefficient for correcting the input quantization step value and determining the output quantization step value for the B picture. A multiplication coefficient as a correction coefficient is stored in the multiplication coefficient two-dimensional LUT 622M with the space change evaluation value h_ACT_MB and the time change evaluation value h_MVbit_MB_B as two variables. An addition coefficient as a correction coefficient is stored in the addition coefficient two-dimensional LUT 622A with the space change evaluation value h_ACT_MB and the time change evaluation value h_MVbit_MB_B as two variables.

  The correction coefficient storage unit 623 stores the correction coefficient for correcting the input quantization step value and determining the output quantization step value for the I picture. The multiplication coefficient is stored in the multiplication coefficient one-dimensional LUT 623M with the spatial change evaluation value h_ACT_MB as one variable. An addition coefficient as a correction coefficient is stored in the addition coefficient one-dimensional LUT 623A with the spatial change evaluation value h_ACT_MB as one variable. For the I picture, the temporal change evaluation value h_MVbit_MB is not a variable of the multiplication coefficient one-dimensional LUT 623M and the addition coefficient one-dimensional LUT 623A.

  Next, description will be made regarding the processing contents in which the second quantization unit 62 determines the output quantization step value by correcting the input quantization step value for a certain macroblock.

  The correction coefficient storage unit 621 receives a spatial change evaluation value h_ACT_MB and a temporal change evaluation value h_MVbit_MB_P for a macroblock having a P picture. The correction coefficient storage unit 621 calculates a multiplication coefficient using the multiplication coefficient two-dimensional LUT 621M. The correction coefficient storage unit 621 calculates the addition coefficient using the addition coefficient two-dimensional LUT 621A. The correction coefficient storage unit 621 outputs the multiplication coefficient and the addition coefficient to the selector 624.

  The correction coefficient storage unit 622 receives the spatial change evaluation value h_ACT_MB and the temporal change evaluation value h_MVbit_MB_B for a macroblock having a B picture. The correction coefficient storage unit 622 calculates a multiplication coefficient using the multiplication coefficient two-dimensional LUT 622M. The correction coefficient storage unit 622 calculates the addition coefficient using the addition coefficient two-dimensional LUT 622A. The correction coefficient storage unit 622 outputs the multiplication coefficient and the addition coefficient to the selector 624.

  The correction coefficient storage unit 623 receives the spatial change evaluation value h_ACT_MB for a macroblock having an I picture. The correction coefficient storage unit 623 calculates a multiplication coefficient using the multiplication coefficient one-dimensional LUT 623M. The correction coefficient storage unit 623 calculates an addition coefficient using the addition coefficient one-dimensional LUT 623A. The correction coefficient storage unit 623 outputs the multiplication coefficient and the addition coefficient to the selector 624.

  FIG. 9A shows the contents of the multiplication coefficient two-dimensional LUTs 621M and 622M. As the time change evaluation value h_MVbit_MB increases, the multiplication coefficient decreases monotonously. As the space change evaluation value h_ACT_MB increases, the multiplication coefficient decreases monotonously. The multiplication coefficients two-dimensional LUTs 621M and 622M may be the same as shown in FIG. 9A or different from those shown in FIG. 9A.

  FIG. 9B is a diagram illustrating the contents of the addition coefficient two-dimensional LUTs 621A and 622A. As the time change evaluation value h_MVbit_MB increases, the addition coefficient decreases monotonously. As the space change evaluation value h_ACT_MB increases, the addition coefficient decreases monotonously. The addition coefficient two-dimensional LUTs 621A and 622A may be the same as shown in FIG. 9B or different from those shown in FIG. 9B.

  FIG. 9C shows the contents of the multiplication coefficient one-dimensional LUT 623M. As the space change evaluation value h_ACT_MB increases, the multiplication coefficient decreases monotonously.

  FIG. 9D shows the contents of the addition coefficient one-dimensional LUT 623A. As the space change evaluation value h_ACT_MB increases, the addition coefficient decreases monotonously.

  The logic circuit unit 625 inputs a picture type P_type_MB for a certain macroblock. The logic circuit portion 625 outputs a switching signal for the selector 624 to the selector 624. FIG. 10 is a flowchart showing the flow of processing in the logic circuit unit 625.

  When the logic circuit unit 625 receives the picture type P_type_MB (step S1), consider that the picture type P_type_MB is a P picture (YES in step S2). The logic circuit unit 625 outputs “0” corresponding to the P picture as the switching signal of the selector 624 (step S3). The selector 624 outputs the multiplication coefficient and the addition coefficient calculated by the multiplication coefficient two-dimensional LUT 621M and the addition coefficient two-dimensional LUT 621A.

  When the logic circuit unit 625 receives the picture type P_type_MB (step S1), consider that the picture type P_type_MB is a B picture (NO in step S2 and YES in step S4). The logic circuit unit 625 outputs “1” corresponding to the B picture as a switching signal of the selector 624 (step S5). The selector 624 outputs the multiplication coefficient and the addition coefficient calculated by the multiplication coefficient two-dimensional LUT 622M and the addition coefficient two-dimensional LUT 622A.

  When the logic circuit unit 625 inputs the picture type P_type_MB (step S1), consider that the picture type P_type_MB is an I picture (NO in step S2, NO in step S4, YES in step S6). The logic circuit unit 625 outputs “2” corresponding to the I picture as the switching signal of the selector 624 (step S7). The selector 624 outputs the multiplication coefficient and the addition coefficient calculated by the multiplication coefficient one-dimensional LUT 623M and the addition coefficient one-dimensional LUT 623A.

  When the logic circuit unit 625 inputs the picture type P_type_MB (step S1), consider that the picture type P_type_MB is an intra macroblock of a P picture or a B picture (NO in step S2, NO in step S4, NO in step S6) ). The logic circuit unit 625 instructs the selector 624 to output the already calculated multiplication coefficient and addition coefficient for the macroblock located in the vicinity of the macroblock (step S8).

  Multiplier 626 multiplies the input quantization step value by the multiplication coefficient output by selector 624. The adder 627 adds the addition coefficient output from the selector 624 to the input quantization step value.

  The selector 628 inputs any one of “0”, “1”, and “2” as a switching signal. When the switching signal is “0”, the selector 628 outputs the uncorrected input quantization step value as the output quantization step value. When the switching signal is “1”, the selector 628 outputs the input quantization step value multiplied by the multiplication coefficient as the output quantization step value. When the switching signal is “2”, the selector 628 outputs the input quantization step value added with the addition coefficient as the output quantization step value.

  For an image region having a large degree of change in the spatial direction and a large degree of change in the time direction, the output quantization step value is set small. The output quantization step value is set to be large for an image region having a small degree of change in the spatial direction and a small degree of change in the time direction.

  That is, for an image region in which a small object moves violently, the image quality can be further improved by assigning a large number of output bits. For an image region in which a large object moves slowly, the image quality is not excessively degraded even if a large number of output bits are not assigned. Thereby, the code amount adjustment of the whole image is also achieved.

  Here, in the first determination unit 61, the output quantization step value is set to be small in an image region where the degree of change in the spatial direction is small. That is, a large number of output bits are assigned to a flat image area. However, in the second determination unit 62, the output quantization step value is set to be small in an image region where the degree of change in the spatial direction is large. That is, a large number of output bits are assigned to an image area including a small object. As described above, the first determination unit 61 and the second determination unit 62 allocate a large number of output bits to image regions having different image characteristics.

  The contents of the multiplication coefficient LUT and the addition coefficient LUT may be fixed without changing over time, and may be rewritten from time to time by the CPU 5. The switching signal to the selector 628 may be fixed without changing with time, and may be updated by the CPU 5 every moment. The range of categories is arbitrary, and the number of dimensions using categorization is arbitrary.

[Processing content of third determining unit]
FIG. 11 is a block diagram illustrating components of the third determination unit 63. The third determination unit 63 inputs “input quantization step value”, “h_ACT_MB”, and “h_SkinTone_MB” at the left end of FIG. 11. The third determining unit 63 outputs an “output quantization step value” at the right end of FIG.

  The “input quantization step value” is a quantization step value corrected by the second determination unit 62 in each macroblock. The “output quantization step value” is a quantization step value corrected by the third determination unit 63 in each macroblock.

  “H_ACT_MB” is a spatial change evaluation value indicating the degree of change in the spatial direction in each macroblock. The space change evaluation value h_ACT_MB is calculated as described below. First, the evaluation value calculation unit 7 inputs a decoded image from the decoder 2. Next, the evaluation value calculation unit 7 calculates the spatial change evaluation value ACT_MB by adding the absolute value of the difference between each pixel value and the average pixel value for each pixel in each macroblock of the decoded image. Next, the evaluation value calculation unit 7 performs categorization by the categorization method shown in FIG.

  “H_SkinTone_MB” is a skin color chromaticity evaluation value indicating a skin color chromaticity in each macro block. The skin color chromaticity evaluation value h_SkinTone_MB is calculated as described below. First, the evaluation value calculation unit 7 inputs a decoded image from the decoder 2. Next, the evaluation value calculation unit 7 calculates the flesh color chromaticity of each pixel based on the color difference signal of each pixel in each macroblock of the decoded image, and measures the number of pixels having a high flesh color chromaticity. The skin color chromaticity evaluation value SkinTone_MB is calculated. Next, the evaluation value calculation unit 7 performs categorization by the categorization method shown in FIG.

  The third determination unit 63 includes a correction coefficient storage unit 631, a multiplier 632, an adder 633, a selector 634, and the like. In particular, the correction coefficient storage unit 631 will be described.

  The correction coefficient storage unit 631 stores the correction coefficient in order to correct the input quantization step value and determine the output quantization step value. A multiplication coefficient as a correction coefficient is stored in the multiplication coefficient two-dimensional LUT 631M with the space change evaluation value h_ACT_MB and the skin color chromaticity evaluation value h_SkinTone_MB as two variables. An addition coefficient as a correction coefficient is stored in the addition coefficient two-dimensional LUT 631A with the space change evaluation value h_ACT_MB and the skin color chromaticity evaluation value h_SkinTone_MB as two variables.

  Next, description will be made regarding the processing contents in which the third determination unit 63 determines the output quantization step value after correcting the input quantization step value for a certain macroblock.

  The correction coefficient storage unit 631 inputs the space change evaluation value h_ACT_MB and the skin color chromaticity evaluation value h_SkinTone_MB for a certain macroblock. The correction coefficient storage unit 631 calculates a multiplication coefficient using the multiplication coefficient two-dimensional LUT 631M. The correction coefficient storage unit 631 calculates an addition coefficient using the addition coefficient two-dimensional LUT 631A.

  FIG. 12A shows the contents of the multiplication coefficient two-dimensional LUT 631M. As the flesh color chromaticity evaluation value h_SkinTone_MB increases, the multiplication coefficient decreases monotonously. As the space change evaluation value h_ACT_MB increases, the multiplication coefficient increases monotonously.

  FIG. 12B is a diagram illustrating the contents of the addition coefficient two-dimensional LUT 631A. As the skin color chromaticity evaluation value h_SkinTone_MB increases, the addition coefficient decreases monotonously. As the space change evaluation value h_ACT_MB increases, the addition coefficient increases monotonously.

  The multiplier 632 multiplies the input quantization step value by the multiplication coefficient output from the multiplication coefficient two-dimensional LUT 631M. The adder 633 adds the addition coefficient output from the addition coefficient two-dimensional LUT 631A to the input quantization step value.

  The selector 634 inputs any one of “0”, “1”, and “2” as a switching signal. When the switching signal is “0”, the selector 634 outputs an uncorrected input quantization step value as the output quantization step value. When the switching signal is “1”, the selector 634 outputs the input quantization step value multiplied by the multiplication coefficient as the output quantization step value. When the switching signal is “2”, the selector 634 outputs the input quantization step value added with the addition coefficient as the output quantization step value.

  The output quantization step value is set to be small for an image region having a large flesh color chromaticity and a small degree of change in the spatial direction. For an image region having a small flesh color chromaticity and a large degree of change in the spatial direction, the output quantization step value is set large.

  In other words, the image quality can be further improved by assigning a large number of output bits to the face image area. For image areas other than the face image area, even if a large number of output bits are not assigned, the image quality is not excessively deteriorated. Thereby, the code amount adjustment of the whole image is also achieved.

  The contents of the multiplication coefficient two-dimensional LUT 631M and the addition coefficient two-dimensional LUT 631A may be fixed without changing over time, or may be rewritten by the CPU 5 every moment. The switching signal to the selector 634 may be fixed without changing with time, and may be updated by the CPU 5 every moment. The range of categories is arbitrary, and the number of dimensions using categorization is arbitrary. The image area in which the output quantization step value is decreased may be not only an image area with high skin color chromaticity but also other image areas with high chromaticity.

[Processing content of the fourth determination unit]
FIG. 13 is a block diagram illustrating components of the fourth determination unit 64. The fourth determination unit 64 inputs “input quantization step value”, “MB_Addr_x”, and “MB_Addr_y” at the left end of FIG. The fourth determination unit 64 outputs an “output quantization step value” at the right end of FIG.

  The “input quantization step value” is a quantization step value corrected by the third determination unit 63 in each macroblock. The “output quantization step value” is a quantization step value of an output image output from the transcoder 1 in each macroblock.

  “MB_Addr_x” and “MB_Addr_y” are coordinate position evaluation values indicating the horizontal and vertical coordinate positions of each macroblock, respectively. The coordinate position evaluation values MB_Addr_x and MB_Addr_y are output from the decoder 2. In the following description of the fourth determination unit 64, for the sake of brevity, it is assumed that the size of one frame in the horizontal direction and the vertical direction is the size of five macroblocks.

  The fourth determination unit 64 includes a correction coefficient storage unit 641, a multiplier 642, an adder 643, a selector 644, and the like. In particular, the correction coefficient storage unit 641 will be described.

  The correction coefficient storage unit 641 stores the correction coefficient in order to correct the input quantization step value and determine the output quantization step value. A multiplication coefficient as a correction coefficient is stored in the multiplication coefficient two-dimensional LUT 641M with coordinate position evaluation values MB_Addr_x and MB_Addr_y as two variables. An addition coefficient as a correction coefficient is stored in the addition coefficient two-dimensional LUT 641A with the coordinate position evaluation values MB_Addr_x and MB_Addr_y as two variables. Reflecting that both the horizontal and vertical sizes of one frame are the size of five macroblocks, the number of matrices of the multiplication coefficient two-dimensional LUT 641M and the addition coefficient two-dimensional LUT 641A are both 5 × 5.

  Next, description will be made regarding the processing contents in which the fourth quantization unit 64 determines the output quantization step value by correcting the input quantization step value for a certain macroblock.

  The correction coefficient storage unit 641 inputs coordinate position evaluation values MB_Addr_x and MB_Addr_y for a certain macroblock. The correction coefficient storage unit 641 calculates a multiplication coefficient using the multiplication coefficient two-dimensional LUT 641M. The correction coefficient storage unit 641 calculates an addition coefficient using the addition coefficient two-dimensional LUT 641A.

  FIG. 14A shows the contents of the multiplication coefficient two-dimensional LUT 641M. As the target macroblock approaches the center of the image, the multiplication coefficient decreases monotonously.

  FIG. 14B shows the contents of the addition coefficient two-dimensional LUT 641A. As the target macroblock approaches the center of the image, the addition coefficient decreases monotonously.

  The multiplier 642 multiplies the input quantization step value by the multiplication coefficient output by the multiplication coefficient two-dimensional LUT 641M. The adder 643 adds the addition coefficient output from the addition coefficient two-dimensional LUT 641A to the input quantization step value.

  The selector 644 inputs any one of “0”, “1”, and “2” as a switching signal. When the switching signal is “0”, the selector 644 outputs the uncorrected input quantization step value as the output quantization step value. When the switching signal is “1”, the selector 644 outputs the input quantization step value multiplied by the multiplication coefficient as the output quantization step value. When the switching signal is “2”, the selector 644 outputs the input quantization step value added with the addition coefficient as the output quantization step value.

  For an image region close to the image center, the output quantization step value is set small. For an image region far from the image center, the output quantization step value is set large.

  In other words, the image quality can be further improved by assigning a large number of output bits to the central image area. For image areas other than the central image area, even if a large number of output bits are not assigned, the image quality is not excessively degraded. Thereby, the code amount adjustment of the whole image is also achieved.

  The contents of the multiplication coefficient two-dimensional LUT 641M and the addition coefficient two-dimensional LUT 641A may be fixed without changing over time, or may be rewritten by the CPU 5 every moment. The switching signal to the selector 644 may be fixed without changing with time, and may be updated by the CPU 5 every moment. The coordinate position evaluation values MB_Addr_x and MB_Addr_y may be categorized. At this time, the range of categories is arbitrary, and the number of dimensions using categorization is arbitrary. The image area for decreasing the output quantization step value may be not only the central image area but also the face image area. At this time, the coordinate position of the face image area may be designated by the evaluation value calculation unit 7.

[Summary of quantization step value determination unit]
The quantization step value determination unit 6 corrects the input quantization step value for each macroblock based on various characteristic evaluation values, and determines an output quantization step value. The transcoder 1 can finely adjust the image quality and the code amount.

  In the quantization step value determination unit 6, the number and arrangement order of the determination units such as the first determination unit 61 are arbitrary. In the determining unit such as the first determining unit 61, the number and combination of variables of the multiplication coefficient LUT and the addition coefficient LUT are arbitrary.

  In the quantization step value determination unit 6, for example, the second determination unit 62, the first determination unit 61, the fourth determination unit 64, and the third determination unit 63 may be arranged in the above order. In the multiplication coefficient LUT and the addition coefficient LUT, for example, a skin color chromaticity evaluation value h_SkinTone_MB, a space change evaluation value h_ACT_MB, and a time change evaluation value h_MVbit_MB may be set as three variables.

{Second Embodiment}
[Components of quantization step value determination unit]
Next, a second embodiment will be described. The quantization step value determination unit 6 is mainly different from the first and second embodiments. FIG. 15 is a block diagram illustrating components of the quantization step value determination unit 6. Comparing the first and second embodiments, the fifth determination unit 65 is disposed downstream of the fourth determination unit 64.

  The fifth determination unit 65 inputs the quantization step value corrected by the fourth determination unit 64 for each macroblock from the fourth determination unit 64 and outputs the quantization step value of the output image output from the transcoder 1. Is output to the encoder 3. The fifth determining unit 65 uses a spatial change evaluation value indicating the degree of change in the spatial direction when correcting the quantization step value corrected by the fourth determining unit 64 for each macroblock.

[Processing content of fifth determining unit]
FIG. 16 is a block diagram showing components of the fifth determining unit 65. The fifth determination unit 65 inputs “input quantization step value”, “ACT_MB”, and “ACT_PIC” at the left end of FIG. The fifth determining unit 65 outputs an “output quantization step value” at the right end of FIG.

  The “input quantization step value” is a quantization step value corrected by the fourth determination unit 64 in each macroblock. The “output quantization step value” is a quantization step value of an output image output from the transcoder 1 in each macroblock.

  “ACT_MB” is a spatial change evaluation value indicating the degree of change in the spatial direction in each macroblock. “ACT_PIC” is a spatial change evaluation average value indicating the average of the spatial change evaluation values ACT_MB of all macroblocks in each frame. The space change evaluation value ACT_MB and the space change evaluation average value ACT_PIC are calculated as described below. First, the evaluation value calculation unit 7 inputs a decoded image from the decoder 2. Next, the evaluation value calculation unit 7 calculates the spatial change evaluation value ACT_MB by adding the absolute value of the difference between each pixel value and the average pixel value for each pixel in each macroblock of the decoded image. Next, the evaluation value calculation unit 7 calculates the spatial change evaluation average value ACT_PIC by calculating the average of the spatial change evaluation values ACT_MB of all macroblocks in each frame of the decoded image.

  The fifth determination unit 65 includes shifters 651 and 652, a subtracter 653, a multiplier 654, a limiter 655, an adder 656, and the like. The processing contents will also be described.

  The shifter 651 receives the space change evaluation value ACT_MB. Then, the shifter 651 performs a shift operation for digit alignment on the space change evaluation value ACT_MB. The shifter 652 inputs the spatial change evaluation average value ACT_PIC. Then, the shifter 652 performs a shift operation for digit alignment on the space change evaluation average value ACT_PIC.

  The subtractor 653 receives the spatial change evaluation value ACT_MB for which the shift operation has been performed and the spatial change evaluation average value ACT_PIC for which the shift operation has been performed. Then, the subtractor 653 subtracts the spatial change evaluation average value ACT_PIC for which the shift calculation has been performed from the spatial change evaluation value ACT_MB for which the shift calculation has been performed. The subtraction value by the subtractor 653 is a value indicating how much the degree of change in the spatial direction in each macroblock varies compared to the average degree of change in the spatial direction in the frame to which each macroblock belongs.

The multiplier 654 receives the subtraction value from the subtracter 653. The multiplier 654 multiplies the subtraction value obtained by the subtracter 653 by an adjustment coefficient r _aq that is a positive number smaller than 1. As a result, the multiplier 654 reduces the fluctuation width of the subtraction value by the subtractor 653.

  The limiter 655 receives the product multiplied by the multiplier 654. Further, the limiter 655 inputs LIMIT_THRES_H as an upper limit threshold and inputs LIMIT_THRES_L as a lower limit threshold. The limiter 655 clips the multiplication value by the multiplier 654 with the upper limit threshold LIMIT_THRES_H, and clips the multiplication value by the multiplier 654 with the lower limit threshold LIMIT_THRES_L. As a result, limiter 655 reduces the amplitude of the product multiplied by multiplier 654.

  The adder 656 receives the input quantization step value and the output value from the limiter 655. Then, the adder 656 determines the output quantization step value in each macroblock by adding the output value from the limiter 655 to the input quantization step value.

  For an image region where the degree of change in the spatial direction is small, the output quantization step value is set small. For an image region having a large degree of change in the spatial direction, the output quantization step value is set large. The output quantization step value will not be too small and not too large.

  That is, for an image region having a small degree of change in the spatial direction, the image quality can be further improved by assigning a large number of output bits. For an image region having a large degree of change in the spatial direction, even if a large number of output bits are not assigned, the image quality is not excessively deteriorated. Thereby, the code amount adjustment of the whole image is also achieved.

The adjustment coefficient r _aq , the upper limit threshold LIMIT_THRES_H, and the lower limit threshold LIMIT_THRES_L may be fixed without changing over time, or may be rewritten by the CPU 5 every moment. In the quantization step value determination unit 6, the number and arrangement order of the determination units such as the first determination unit 61 are arbitrary.

{Third embodiment}
[Components of quantization step value determination unit]
Next, a third embodiment will be described. The quantization step value determination unit 6 is mainly different from the second and third embodiments. FIG. 17 is a block diagram illustrating components of the quantization step value determination unit 6. Compared with the second and third embodiments, the fine adjustment unit 66 is arranged downstream of the fifth determination unit 65.

  The fine adjustment unit 66 inputs the quantization step value corrected by the fifth determination unit 65 from the fifth determination unit 65 for each macroblock, and calculates the quantization step value of the output image output from the transcoder 1. Output to the encoder 3. The fine adjustment unit 66 uses the multiplication fine adjustment coefficient and the addition fine adjustment coefficient for fine adjustment of the quantization step value corrected by the fifth determination unit 65 for each macroblock.

[Details of fine adjustment processing]
FIG. 18 is a block diagram illustrating components of the fine adjustment unit 66. The fine adjustment unit 66 inputs “input quantization step value” and “P_type_MB” at the left end of FIG. The fine adjustment unit 66 outputs an “output quantization step value” at the right end of FIG.

  The “input quantization step value” is a quantization step value corrected by the fifth determination unit 65 in each macroblock. The “output quantization step value” is a quantization step value of an output image output from the transcoder 1 in each macroblock.

  “P_type_MB” is a parameter indicating the picture type in each macroblock. Examples of picture types include intra macroblocks in I pictures, P pictures, B pictures, P pictures, or B pictures.

  The fine adjustment unit 66 includes a multiplier 661, a selector 662, an adder 663, a selector 664, and the like. The processing contents will also be described.

  The multiplier 661 receives the input quantization step value. Then, the multiplier 661 multiplies the input quantization step value by a multiplication fine adjustment coefficient. Here, the fine multiplication adjustment coefficient is determined as described below.

The selector 662 inputs the picture type P_type_MB as a switching signal. When the picture type P_type_MB is an intra macroblock of an I picture, P picture, B picture, P picture, or B picture, the selector 662 uses α _I , α _P , α _B , α _I− as the multiplication fine adjustment coefficients, respectively _{. PB} is output.

  The adder 663 receives the multiplication value obtained by the multiplier 661. The adder 663 adds an addition fine adjustment coefficient to the multiplication value obtained by the multiplier 661. Here, the addition fine adjustment coefficient is determined as described below.

The selector 664 receives the picture type P_type_MB as a switching signal. When the picture type P_type_MB is an I-picture, P-picture, B-picture, P-picture, or B-picture intra macroblock, the selector 664 uses β _I , β _P , β _B , β _{I- PB} is output.

From the first determination unit 61 to the fifth determination unit 65, the output quantization step value is corrected, and in the fine adjustment unit 66, the output quantization step value is finely adjusted. The multiplication fine adjustment coefficients α _I , α _P , α _B , α _I-PB , and the addition fine adjustment coefficients β _I , β _P , β _B , β _I-PB may be fixed without changing over time. The CPU 5 may rewrite the data every moment. In the quantization step value determination unit 6, the number and arrangement order of the determination unit such as the first determination unit 61 and the fine adjustment unit such as the fine adjustment unit 66 are arbitrary.

1 Transcoder 2 Decoder 3 Encoder 4 Processing Setting Unit 5 CPU
6 quantization step value determination unit 7 evaluation value calculation unit 8 image memories 61, 61A, 61B first determination unit 62 second determination unit 63 third determination unit 64 fourth determination unit 65 fifth determination unit 66 fine adjustment unit

Claims (25)

A decoder that decodes an input image to generate a decoded image;
An encoder that encodes the decoded image to generate an output image;
A transcoder comprising:
A spatial change evaluation value calculation unit that calculates a spatial change evaluation value indicating the degree of change in the spatial direction in each macroblock;
A time change evaluation value calculation unit for calculating a time change evaluation value indicating the degree of change in the time direction in each macroblock;
Based on the space change evaluation value and the time change evaluation value, a quantization coefficient is used to correct an input quantization step value of each macroblock and determine an output quantization step value of each macroblock. A step value determination unit;
A transcoder comprising: The transcoder of claim 1 .
The quantization step value determination unit determines an output quantization step value to be smaller as the spatial change evaluation value is larger, and determines an output quantization step value to be smaller as the temporal change evaluation value is larger. Transcoder. The transcoder according to claim 1 or 2 ,
The transcoder, wherein the quantization step value determination unit determines an output quantization step value based on a picture type of each macroblock. The transcoder according to any one of claims 1 to 3,
The spatial change evaluation value is
In each macroblock of the decoded image, the difference absolute value sum of each pixel value and the average pixel value for each pixel is added,
Transcoder characterized by including. The transcoder according to any one of claims 1 to 4,
The time change evaluation value is
The magnitude of the motion vector in each macroblock of the input image,
Transcoder characterized by including. The transcoder according to any one of claims 1 to 4,
The time change evaluation value is
Code amount of motion vector in each macroblock of the input image,
Transcoder characterized by including. The transcoder according to any one of claims 1 to 6,
The quantization step value determination unit
A correction coefficient storage unit that stores the correction coefficient using the space change evaluation value and the time change evaluation value as variables,
Transcoder characterized by including. A decoder that decodes an input image to generate a decoded image;
An encoder that encodes the decoded image to generate an output image;
A transcoder comprising:
A spatial change evaluation value calculation unit that calculates a spatial change evaluation value indicating the degree of change in the spatial direction in each macroblock;
A predetermined chromaticity evaluation value calculating unit that calculates a predetermined chromaticity evaluation value indicating a predetermined chromaticity in each macroblock;
Based on the spatial change evaluation value and the predetermined chromaticity evaluation value, the input quantization step value of each macroblock is corrected by using a correction coefficient, and the output quantization step value of each macroblock is determined. A quantization step value determination unit;
A transcoder comprising: The transcoder according to claim 8 , wherein
The predetermined chromaticity evaluation value is:
Skin color chromaticity evaluation value indicating the skin color chromaticity in each macroblock of the decoded image,
Including
The quantization step value determination unit determines an output quantization step value to be smaller as the spatial change evaluation value is smaller, and determines an output quantization step value to be smaller as the skin color chromaticity evaluation value is larger. A characteristic transcoder. The transcoder according to claim 8 or 9 ,
The spatial change evaluation value is
In each macroblock of the decoded image, the difference absolute value sum of each pixel value and the average pixel value for each pixel is added,
Transcoder characterized by including. The transcoder according to any one of claims 8 to 10 ,
The quantization step value determination unit
A correction coefficient storage unit that stores the correction coefficient using the space change evaluation value and the predetermined chromaticity evaluation value as variables,
Transcoder characterized by including. A decoder that decodes an input image to generate a decoded image;
An encoder that encodes the decoded image to generate an output image;
A transcoder comprising:
A coordinate position evaluation value calculation unit for calculating a coordinate position evaluation value indicating the position of each macroblock;
A quantization step value determining unit that corrects an input quantization step value of each macroblock by using a correction coefficient based on the coordinate position evaluation value and determines an output quantization step value of each macroblock; ,
A transcoder comprising: The transcoder according to claim 12 , wherein
The quantization step value determining unit determines the output quantization step value to be smaller as the position of each macroblock is closer to the center of the image. The transcoder according to claim 12 or claim 13 ,
The quantization step value determination unit
A correction coefficient storage unit that stores the correction coefficient using the coordinate position evaluation value as a variable,
Transcoder characterized by including. The transcoder according to any one of claims 7 , 11 , and 14 ,
A correction coefficient that is updated and written to the correction coefficient storage unit every moment based on temporal changes of the input bit rate of the input image, the output bit rate of the output image, and the target bit rate of the output image Update writing section,
A transcoder comprising: The transcoder according to any one of claims 1 to 15 , further comprising:
An evaluation value classification unit that performs a predetermined number of classifications on the evaluation values in each macroblock,
With
The quantization step value determination unit determines an output quantization step value based on an evaluation value that has been subjected to a predetermined number of divisions. The transcoder of claim 16 .
A transcoder characterized in that the evaluation value ranges in each section are equal. The transcoder of claim 16 .
Transcoders characterized by different evaluation value ranges in each category. The transcoder according to any one of claims 1 to 14,
The correction factor is
A multiplication correction factor by which the input quantization step value is multiplied in order to determine the output quantization step value;
In order to determine the output quantization step value, an addition correction coefficient added to the input quantization step value;
Transcoder characterized by including. The transcoder of claim 19 , further comprising:
A quantization step value obtained by multiplying the input quantization step value by the multiplication correction coefficient, a quantization step value obtained by adding the addition correction coefficient to the input quantization step value, and an input quantization step value obtained by the multiplication correction coefficient. Output quantization step value is selected as the output quantization step value from among the quantization step values that are not multiplied by and the addition correction coefficient is not added to the input quantization step value Step value selector,
A transcoder comprising: The transcoder according to any one of claims 1 to 20 ,
The quantization step value determination unit
A fine adjustment unit that calculates an output quantization step value finely adjusted by a fine adjustment coefficient with respect to the output quantization step value corrected by the correction coefficient,
Transcoder characterized by including. The transcoder of claim 21 .
The fine adjustment factor is
A multiplication fine adjustment coefficient to be multiplied by the output quantization step value corrected by the correction coefficient;
Transcoder characterized by including. The transcoder according to claim 21 or claim 22 ,
The fine adjustment factor is
An addition fine adjustment coefficient to be added to the output quantization step value corrected by the correction coefficient;
Transcoder characterized by including. 24. The transcoder according to any one of claims 21 to 23 , wherein:
The transcoder, wherein the fine adjustment coefficient is set based on a picture type of each macroblock.
25. The transcoder according to any one of claims 21 to 24 , wherein:
The fine adjustment coefficient is updated from time to time based on temporal changes of the input bit rate of the input image, the output bit rate of the output image, and the target bit rate of the output image. .

Images