JP2011109216A - Encoding method, and encoding program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To solve the following problem: when a feature of an image significantly varies like a scene change in moving image encoding, large increase of a calculation volume occurs. <P>SOLUTION: The encoding method includes: a correlation calculation step (S201) of calculating a correlation value related to an image of an encoding object; a quantization parameter determination step (S202) of determining a quantization parameter in performing encoding by using the correlation value calculated in the correlation calculation step (S201) based on a correspondence relation between the correlation value and a quantization parameter in performing encoding; and an encoding step (S203) of encoding the image using the quantization parameter determined in the quantization parameter determination step (S202). <P>COPYRIGHT: (C)2011,JPO&INPIT

Description

  The present invention relates to moving image encoding, and more particularly to a moving image encoding method and encoding program that enable efficient parallel processing.

  In recent years, digitalization of AV information has progressed and devices capable of digitizing and handling video signals are becoming widespread. Since a video signal has an enormous amount of information, encoding is generally performed while reducing the amount of information in consideration of recording capacity and transmission efficiency. As a video signal encoding technique, H.264 is used. An international standard called H.264 has been established. The H.264 standard provides approximately twice the compression efficiency compared to other existing coding technologies. Therefore, H.I. The H.264 standard is an encoding method that is very suitable for recorder recording, Internet streaming, and the like.

  In a video recorder that encodes a video signal and records it on a magnetic tape or a semiconductor memory, the amount of generated code in one frame or one field (hereinafter referred to as a picture) is constant (fixed length) in consideration of editing, variable speed playback, etc. It is desirable that As a conventional technique for keeping the generated code amount at a fixed length, Patent Document 1 selects an arbitrary Qp as a candidate from all the quantization parameters (hereinafter referred to as Qp) of the system, and actually A technique is disclosed in which encoding is performed using a candidate Qp and a quantization parameter that realizes a maximum code amount that does not exceed a specific target code amount is selected. However, the technique disclosed in Patent Document 1 requires a very large amount of calculation. A technique shown in FIG. 7 is widely known as a conventional technique for reducing the amount of calculation.

  In FIG. 7, step S600 is performed only once for the sequence. Steps S602 to S606 are repeatedly executed for each frame image or field image.

  First, in step S600, (the maximum Qp that the system has) / 2 is set as the initial value of Qp_prev. In this example, Qp_prev = 31. In step S602, Qp = Qp_prev, and in step S603, the picture is encoded using the quantization parameter Qp. After the encoding of the encoding target picture is completed, it is determined in step S605 whether or not the picture code amount is equal to or less than the target code amount.

  If it is determined that the picture code amount exceeds the target code amount, Qp = Qp + 1 is set in step S604, and encoding is performed again in step S603. Thereafter, the same processing is repeated until it is determined in step S605 that the picture code amount is equal to or less than the target code amount.

  If it is determined that the picture code amount is equal to or less than the target code amount, Qp_prev = Qp is set in step S606, and the encoding process of the picture is ended.

  In this method, when the generated code amount is remarkably small with respect to the target code amount, a branch in which re-encoding is performed in step 603 with Qp = Qp−1 can be considered, but this is omitted for convenience of explanation.

JP 7-322203 A

  However, these conventional techniques have a problem that the amount of calculation for determining the final Qp is large.

  In the prior art, fixed-length encoding is guaranteed by repeating the encoding process by changing Qp until the picture code amount falls within the fixed code amount. Since the picture starts to be encoded using the final Qp of the immediately preceding picture, it is unlikely that the code amount will fluctuate greatly in a normal image with a high temporal correlation, and the amount of computation until the final Qp determination is reduced. I can expect. However, when the image characteristics change greatly as in the case of a scene change, the adverse effect of repeating the encoding until the picture code amount falls within a fixed code amount becomes significant.

  For example, it is assumed that Qp_prev = 26, the target picture is an image after a scene change, and the optimum Qp is 35. This means that the generated code amount falls within the fixed code amount only when Qp becomes 35 or more. In the conventional example, encoding is performed with Qp = 26 in the first encoding, but since the generated code amount exceeds the target code amount, it is updated to Qp = 27 and the second encoding is performed. In the second encoding, encoding is performed with Qp = 27. However, since the generated code amount exceeds the target code amount, it is updated to Qp = 28 and the third encoding is performed. Thereafter, encoding is repeated until Qp = 35, and a stream of the picture is generated. In this case, the encoding is repeated 10 times to encode one picture.

  From the above, the conventional technique has a problem that a large increase in the amount of calculation occurs when the image characteristics change greatly, such as a scene change.

  In order to solve the above conventional problems, an encoding method of the present invention is an encoding method for encoding an image, a correlation calculating step for calculating a correlation value for an image to be encoded, a correlation value and a code A quantization parameter determining step for determining a quantization parameter for encoding using the correlation value calculated in the correlation calculating step based on a correspondence relationship with a quantization parameter for encoding; An encoding step of encoding an image using the quantization parameter determined in the encoding parameter determination step.

  According to the encoding method of the present invention, an increase in the amount of calculation can be suppressed as much as possible even when the statistical characteristics of an image such as a scene change change greatly.

A block diagram showing the configuration of a PC on which the encoding program according to Embodiment 1 operates Flowchart showing a processing flow of the encoding method according to the first embodiment. The figure which shows the relationship between the correlation value which concerns on Embodiment 1, and a quantization parameter The figure which shows the slice division which concerns on Embodiment 1,2. The flowchart which shows the flow of a process of the encoding method which concerns on Embodiment 2. FIG. The figure explaining the change of the quantization parameter which concerns on Embodiment 2. FIG. The flowchart which shows the flow of a process of the conventional encoding method

  Hereinafter, embodiments will be described with reference to the drawings.

(Embodiment 1)
FIG. 1 is a block diagram showing a configuration of a PC (personal computer) 100 on which an encoding program for executing an encoding method according to Embodiment 1 operates. In FIG. 1, a CPU 101 is a system control unit and controls the entire apparatus. The ROM 102 is a memory that stores an encoding program that runs on the CPU 101. The CPU 101 reads the encoding program stored in the ROM 102 and operates the read encoding program. The ROM 102 stores other control programs and various fixed data. The RAM 103 is composed of SRAM, DRAM, etc., and is a memory for storing program control variables. Various setting parameters and various work buffers are also stored in the RAM 103.

  The operation unit 104 includes a keyboard, a mouse, a touch panel, and the like, and is an instruction input device for an operator to perform various input operations. The display unit 105 is a display device that displays various types of information including images.

  The recording medium 107 stores image data to be encoded, and stores image data generated as a result of encoding. The recording medium 107 is composed of a hard disk, a semiconductor memory card, or the like. The recording medium 107 may be built in the PC 100 or detachable from the PC 100.

  The recording medium control unit 106 is a control unit that controls reading of various information including image data recorded on the recording medium 107 and writing of information to the recording medium 107. The recording medium control unit 106 reads out the image data to be encoded recorded on the recording medium 107 based on the control of the CPU 101, and writes the encoded image data into the recording medium 107.

  The external i / f unit 108 is an interface unit that exchanges various types of information including image data with an external device. The image data to be encoded is input from an external device via the external i / f unit 108 in addition to the form stored in the recording medium 107. The encoded image data may be output to an external device via the external i / f unit 108 in addition to the form stored in the recording medium 107. As the external i / f unit 108, for example, SDI (Serial Digital Interface), LAN, or the like is used.

  Next, the operation of the moving picture coding method according to Embodiment 1 will be described using the flowchart shown in FIG.

  In the figure, first, in step S201, a correlation value val related to a picture to be encoded is detected. The correlation value is desirably an index indicating the complexity of the encoding target picture. In the first embodiment, the sum of absolute values of pixels of the difference image calculated by motion compensation is used. In addition, orthogonal transformation may be performed on the difference image, and the absolute value sum of the transform coefficients may be used as the correlation value.

  In step S202, Qp corresponding to the correlation value val is set. Subsequently, in step S203, the picture is encoded using the quantization parameter Qp. After the encoding of the encoding target picture is completed, it is determined in step S204 whether the picture code amount is equal to or less than the target code amount.

  If it is determined that the picture code amount exceeds the target code amount, Qp = Qp + 1 is set in step S205, and encoding is performed again in step S203. Thereafter, the same processing is repeated until it is determined in step S204 that the picture code amount is equal to or less than the target code amount.

  When it is determined that the picture code amount is equal to or less than the target code amount, the encoding process of the picture is ended.

  The determination of Qp in accordance with the correlation value val in step S202 will be described in more detail with reference to FIG.

  The correlation value val is a value calculated with respect to a signal to be encoded. In intra-frame encoding, the sum of absolute values of pixel values of a difference image after intra prediction. In inter-frame encoding, a difference after motion compensation. The absolute value sum of the pixel values of the image. The pixel value distribution of the image is wide, and there is a large variation depending on the image, but the pixel value of the difference image is distributed in a very narrow range centered on zero called the Laplace distribution, and the variation due to the image is also small. The absolute value sum of the pixel values of the difference image to be encoded indicates the complexity of the image in encoding, and a certain proportional relationship can be seen with respect to the generated code amount after quantization. H. In H.264, when the quantization parameter Qp increases by 6, the quantization step is doubled. When the quantization step is doubled, the generated code amount is considered to be ½ times in the entire image although there is a local error. For these reasons, there is a certain relationship between Qp, the sum of absolute values of the pixel values of the difference image, which is an index as the complexity of the encoding target image, and the generated code amount.

  Now, assuming that the generated code amount is a fixed value, the relationship between Qp and the correlation value val is defined as shown in FIG. As described above, since the pixel values of the difference image have a specific distribution regardless of the image, the values of val_t and Qp_t can be obtained from the test image if the quantization matrix to be used and the target generated code amount are determined. It can be calculated by actually taking statistics of Qp and the amount of generated code. That is, val_t and Qp_t can be set in advance before the start of encoding. For example, H. When a H.264 standard matrix is used, a target generated code amount of a picture is 1000 bits, and val_t is 30000, Qp_t = 26 is set. If the starting points val_t and Qp_t are determined, a table of correlation values val and Qp can be generated from the relationship shown in FIG.

  Therefore, the values of val_t and Qp_t are determined based on the target generated code amount determined at the time of encoding, and then a table representing the relationship shown in FIG. 3 is expressed using the relationship between the quantization parameter and the generated code amount. Create in advance before starting.

  In step S202, Qp corresponding to the correlation value val is determined using a table created in advance.

  In the first embodiment, the table representing the relationship shown in FIG. 3 is created. However, a relational expression representing the relation between the correlation value val and the quantization parameter Qp is derived in advance, and the correlation value is based on the relational expression. Qp corresponding to val may be determined.

  The encoding method according to the first embodiment is very effective even when an image is divided into a plurality of regions and encoding is performed in parallel for each region.

  For example, the encoding target picture is divided into four areas 310 to 313 as shown in FIG. 4, and each area is encoded in parallel in step S <b> 203, so that the encoding speed can be greatly increased.

  As described above, in the encoding method of the first embodiment, the correlation value of the image is obtained from the absolute value sum of the pixels of the difference image calculated by motion compensation, and the correlation is based on the target generated code amount at the time of encoding. By deriving the relationship between the value and the quantization parameter, determining the quantization parameter using the calculated correlation value based on the derived relationship, and performing the encoding using the quantization parameter The quantization parameter used sometimes can be set appropriately, and an increase in the amount of calculation can be suppressed as much as possible.

(Embodiment 2)
The configuration of the PC that executes the encoding method according to the second embodiment is the same as that of the first embodiment, and a description thereof will be omitted.

  The operation of the moving picture coding method according to Embodiment 2 will be described using the flowchart shown in FIG.

  In the figure, step S400 is executed only once for the sequence. Steps S402 to S409 are repeatedly executed for each frame image or field image.

  First, in step S400, (the maximum Qp that the system has) / 2 is set as the initial value of Qp_init and Qp_prev. In this example, Qp_init = Qp_prev = 31. In step S402, a correlation value val between the encoding target picture and the picture encoded immediately before is detected. The correlation value is preferably an index indicating the correlation between the current picture to be coded and the picture coded immediately before. In the second embodiment, the difference between the picture and the picture coded immediately before is desirable. Use the sum of absolute values of the image. In addition, orthogonal transformation may be performed on the difference image, and the absolute value sum of the transform coefficients may be used as the correlation value.

  In step S403, it is determined whether or not the correlation value val is equal to or less than a specific threshold value. If it is determined that val is equal to or less than a specific threshold, the picture is considered to have a high correlation with the picture coded immediately before, and the optimum Qp of the picture is Qp_prev which is the Qp of the picture coded immediately before. Expected to be close to. Accordingly, in step S404, Qp = Qp_prev is set.

  On the other hand, if it is determined that the correlation value val is greater than a specific threshold, the picture is considered to have a low correlation with the picture coded immediately before, and the optimum Qp of the picture is a value deviating from Qp_prev. Expected to take. Thereby, in step S405, Qp = Qp_init is set.

  Subsequently, in step S406, the picture is encoded using the quantization parameter Qp. After the encoding of the encoding target picture is completed, it is determined in step S407 whether or not the picture code amount is equal to or less than the target code amount.

  If it is determined that the picture code amount exceeds the target code amount, Qp = Qp + 1 is set in step S408, and encoding is performed again in step S406. Thereafter, the same processing is repeated until it is determined in step S407 that the picture code amount is equal to or less than the target code amount.

  If it is determined that the picture code amount is equal to or less than the target code amount, Qp_prev = Qp is set in step S409, and the encoding process of the picture is ended.

  The determination of Qp according to the correlation value val in steps S403 to S405 will be described in more detail with reference to FIG.

  In FIG. 6, picture 0, picture 1, and picture 2 are part of a continuous moving image, and a scene change has occurred in picture 2. The optimum Qp for pictures 0, 1 and 2 were 26, 26 and 35, respectively. This indicates that Qp which gives the maximum code amount not exceeding the target code amount for picture 0 is 26.

  Assume that the correlation value val of each picture is 1000 for picture 0, 1000 for picture 1, 40000 for picture 2, and a threshold value 20000. At this time, step S403 performs the following determination.

  First, since the correlation value val for picture 0 is 1000 and is smaller than the threshold value 20000, it is considered that the correlation with the picture coded immediately before is high. As a result, Qp of picture 0 is set to 26, which is the value of Qp_prev.

  Subsequently, for picture 1, since the correlation value val is 1000 and is smaller than the threshold value 20000, it is considered that the correlation with the picture coded immediately before is high. As a result, Qp of picture 1 is set to 26, which is the value of Qp_prev.

  Subsequently, for picture 2, since the correlation value val is 40000 and is larger than the threshold value 20000, it is considered that the correlation with the picture coded immediately before is low. As a result, Qp of picture 2 is set to 31, which is the value of Qp_init.

  On the other hand, in the conventional example, the Qp of the picture encoded immediately before is used regardless of the correlation of the images, so that the Qp of each picture is 26. In the conventional example and the second embodiment, when each picture is encoded, the number of encodings required until the end of the picture encoding is compared.

  First, for picture 0, the optimum Qp is 26. In the conventional example, encoding is performed in step S603 with Qp = 26. In step S605, the picture generation code amount is compared with the target code amount, it is determined that the picture generation code amount is equal to or less than the target code amount, and the encoding is terminated. The number of encodings required until the end of encoding the picture is one. In the second embodiment, encoding is performed in step S406 with Qp = 26. In step S407, the picture generation code amount is compared with the target code amount, it is determined that the picture generation code amount is equal to or less than the target code amount, and the encoding is terminated. The number of encodings required until the end of encoding the picture is one.

  Subsequently, for picture 1, the optimum Qp is 26. In the conventional example, encoding is performed in step S603 with Qp = 26. In step S605, the picture generation code amount is compared with the target code amount, it is determined that the picture generation code amount is equal to or less than the target code amount, and the encoding is terminated. The number of encodings required until the end of encoding the picture is one. In the second embodiment, encoding is performed in step S406 with Qp = 26. In step S407, the picture generation code amount is compared with the target code amount, it is determined that the picture generation code amount is equal to or less than the target code amount, and the encoding is terminated. The number of encodings required until the end of encoding the picture is one.

  Subsequently, for picture 2, the optimum Qp is 35. In the conventional example, encoding is performed in step S603 with Qp = 26. In step S605, the picture generation code amount is compared with the target code amount, and it is determined that the picture generation code amount is not less than or equal to the target code amount. In step S604, Qp = 27 is updated. Encoding is performed in step S603 with Qp = 27. Thereafter, the same processing is performed, and the encoding is finished at Qp = 35. The number of encodings required to complete the encoding of the picture is 10 times. In the second embodiment, encoding is performed in step S406 with Qp = 31. In step S407, the picture generation code amount is compared with the target code amount, and it is determined that the picture generation code amount is not less than or equal to the target code amount. In step S408, Qp = 27 is updated. Encoding is performed in step S406 with Qp = 27. Thereafter, the same processing is performed, and the encoding is finished at Qp = 35. The number of encodings required until the end of encoding the picture is five.

  The encoding method of the second embodiment is very effective even when an image is divided into a plurality of regions and encoding is performed in parallel for each region.

  For example, the picture to be encoded is divided into four areas 310 to 313 as shown in FIG. 4, and each area is encoded in parallel in step S406, so that the encoding speed can be significantly increased.

  As described above, according to the encoding method of the second embodiment, the correlation value is calculated from the sum of absolute values of the difference image between the encoding target picture and the picture encoded immediately before, and the correlation value has a threshold value. If this is exceeded, by using the initial value as the quantization parameter, it is possible to reduce the number of times of encoding to make the generated code amount of the encoding target picture equal to or less than the target code amount. Even when the characteristics change greatly, an increase in the amount of calculation can be suppressed as much as possible.

  According to the encoding method according to the present embodiment, an increase in the amount of calculation can be suppressed as much as possible even when the statistical characteristics of an image such as a scene change greatly change. Useful for.

100 PC
101 CPU
102 ROM
103 RAM
104 Operation unit 105 Display unit 106 Recording medium control unit 107 Recording medium 108 External i / f unit

Claims (10)

An encoding method for encoding an image, comprising:
A correlation calculating step of calculating a correlation value related to an image to be encoded;
Based on the correlation between the correlation value and the quantization parameter at the time of encoding, the quantization parameter determination step for determining the quantization parameter at the time of encoding using the correlation value calculated in the correlation calculation step When,
An encoding step of encoding an image using the quantization parameter determined in the quantization parameter determination step;
An encoding method including: An encoding method for encoding an image, comprising:
A correlation calculating step of calculating a correlation value related to an image to be encoded;
When the correlation value calculated in the correlation calculation step exceeds a predetermined value, the initial value is used as the quantization parameter, and when the correlation value does not exceed the predetermined value, the quantization parameter used when encoding immediately before is used. A quantization parameter determination step;
An encoding step of encoding an image using the quantization parameter determined in the quantization parameter determination step;
An encoding method including: The correlation detection step includes
The encoding method according to claim 1, wherein a correlation value is calculated based on an absolute value of an entire image of difference pixels between an image to be encoded and an image encoded immediately before. The correlation detection step includes
The encoding method according to claim 1, wherein a correlation value is calculated based on an absolute value of a pixel value of a residual image after motion compensation. The encoding step includes
The encoding method according to any one of claims 1 to 4, wherein an image to be encoded is divided into a plurality of regions, and encoding is performed in parallel for each region. A computer-executable encoding program for encoding an image,
A correlation calculating step of calculating a correlation value related to an image to be encoded;
Based on the correlation between the correlation value and the quantization parameter at the time of encoding, the quantization parameter determination step for determining the quantization parameter at the time of encoding using the correlation value calculated in the correlation calculation step When,
An encoding step of encoding an image using the quantization parameter determined in the quantization parameter determination step;
An encoding program including A computer-executable encoding program for encoding an image signal,
A correlation calculating step of calculating a correlation value related to an image to be encoded;
When the correlation value calculated in the correlation calculation step exceeds a predetermined value, the initial value is used as the quantization parameter, and when the correlation value does not exceed the predetermined value, the quantization parameter used when encoding immediately before is used. A quantization parameter determination step;
An encoding step of encoding an image using the quantization parameter determined in the quantization parameter determination step;
An encoding program including The correlation detection step includes
The encoding program according to any one of claims 6 and 7, wherein a correlation value is calculated based on an absolute value of an entire image of difference pixels between an image to be encoded and an image encoded immediately before. The correlation detection step includes
The encoding program according to any one of claims 6 and 7, wherein a correlation value is calculated based on an absolute value of a pixel value of a residual image after motion compensation. The encoding step includes
The encoding program according to any one of claims 6 to 9, wherein an encoding target image is divided into a plurality of regions, and encoding is performed in parallel for each region.

Images