JP2019097128A - Video image coding parameter adjustment device, video image coding parameter adjustment method, and program - Google Patents

Abstract

To simply and precisely adjust a video image coding parameter.SOLUTION: A control part 104 of a video image coding device 101 acquires a different video image coding parameter value used when coding each scene structuring a test video image loop-reproduced at a plurality of times. An RD cost calculation part 105 calculates an RD cost for coding performed to the video image of each scene by a video image coding part 103 by using the video image coding parameter value acquired by the control part 104. An optimal parameter determination part 106 determines the video image coding parameter value suitable for the coding on the basis of the association with the RD cost and the video image coding parameter value.SELECTED DRAWING: Figure 1

Description

  The present invention relates to a video coding parameter adjustment device, a video coding parameter adjustment method, and a program.

  Usually, in a video coding system or a video coding apparatus, a plurality of video coding parameters (hereinafter also referred to as “coding parameters”) which determine coding efficiency exist. Adjusting these coding parameters in accordance with the coding conditions and the characteristics of the input video leads to improvement in coding efficiency and image quality at the time of compression. Therefore, adjustment of coding parameters is an important issue. For example, in a video encoding apparatus that performs real-time encoding processing, a plurality of setting data or firmware in which the values of encoding parameters to be adjusted are gradually changed are prepared, and the same test video is encoded using these. The value of the coding parameter is determined based on the superiority or inferiority of the coding result at the time of coding.

  Specifically, the code amount R of each picture is determined from the log at the time of encoding or the collected encoded stream. Furthermore, the difference between the decoded video and the original image is determined by a dedicated device or software processing, and the encoding distortion amount D of each picture is determined. Based on these, an RD cost which is an index obtained by linearly combining the code amount R and the coding distortion amount D, and a value of a coding parameter which minimizes D + λR are determined as an optimum parameter value (for example, non-patent) Reference 1). Here, λ is a value called a Lagrange coefficient, which corresponds to (slope of tangent of RD characteristic curve) × (−1).

Supervised by Kei Okubo, "H. 265 / HEVC Textbook", pp. 232-239, published Impress Japan, 2013

  In general, the setting value and the optimum value of most of the encoding parameters differ depending on the magnitude of the quantization parameter. Therefore, instead of normal encoding with fixed bit rate control, encoding is performed with fixed quantization parameter settings, and the RD cost is obtained from the code amount R and the encoding distortion amount D at that time, and the optimal parameter is determined. judge. Thereby, the optimum value of the parameter corresponding to the quantization parameter can be determined. In addition, the optimal parameter value also differs depending on the type of picture, for example, a picture type such as I picture, P picture or B picture, or Temporal_ID representing a hierarchy of reference relationships in a GOP (Group of Picture) (for example, non-patent Reference 1). Therefore, it is necessary to adjust RD by aggregating RD cost for each picture type and each Temporal_ID. In these operations, there is a problem that it takes much time and labor for operations such as video input, encoding, stream collection, acquisition of decoded video, aggregation and analysis of various logs.

  Further, as a method of improving the operation of the method as described above, there is also a method of preparing software for simulating a video encoding apparatus or a video encoding system and adjusting parameters. For example, a large number of setting files with different parameter settings are prepared. Then, encoding using the setting files is performed in parallel or sequentially by a computer, and software automatically executes until RD costs are totaled. This enables efficient adjustment. However, this method has the problem that it requires a lot of computer resources, the problem that the execution speed of the simulation software is slow, and the problem that the accuracy is low due to the mismatch between the operation of the simulation software and the video encoding device .

  In view of the above circumstances, the present invention has an object of providing a video coding parameter adjusting device, a video coding parameter adjusting method, and a program capable of performing adjustment of video coding parameters accurately and easily.

  According to an aspect of the present invention, there is provided an acquisition unit for acquiring different video coding parameter values used in coding each of a plurality of small video included in a large video, and the video coding parameter for each small video. A video code suitable for coding based on an evaluation unit which performs coding evaluation performed on the small video using a value based on a predetermined index, and a relation between the evaluation and the video coding parameter value And a determination unit that determines an equalization parameter value.

  One embodiment of the present invention is the above-described video coding parameter adjustment device, wherein the large video includes the small video to be loop-reproduced, and the head of the loop is detected based on the video feature of the large video. A detection unit, a change unit for changing the video coding parameter value used when encoding the small video at the head of the loop detected by the detection unit, and the loop detected by the detection unit And an encoding unit for initializing the GOP (Group of Picture) phase at the beginning of the group, and encoding the small image from the beginning of the GOP using the image encoding parameter value changed by the changing unit for each loop. A video coding parameter adjustment device.

  One embodiment of the present invention is the above-described video coding parameter adjustment device, wherein the evaluation unit performs, for each picture type, a code performed on a picture included in the small video using the video coding parameter value. Evaluation is performed on the basis of the predetermined index, and the determination unit determines an image suitable for encoding for each of the picture types based on the relation between the evaluation for each of the picture types and the image encoding parameter value Determine coding parameter values.

  One aspect of the present invention is the above-described video coding parameter adjustment device, wherein the large video includes the small video to be loop-reproduced, and the predetermined index includes a generated code amount R and a coding degradation D. The RD unit is a linear combination, and the evaluation unit encodes the small image of the two loops included in the large image using quantization parameter values before and after the quantization parameter value to be adjusted. The slope of the RD characteristic curve is determined based on the generated code amount R and the coding deterioration D when performed, and the Lagrange coefficient used to calculate the RD cost is determined based on the determined slope.

  One embodiment of the present invention is the above-described video coding parameter adjustment device, wherein the large video includes a plurality of small video to be loop reproduced, and the detection unit is based on the video feature amount of the large video. And the encoding unit initializes a GOP (Group of Picture) phase at the beginning of the small image detected by the detection unit, and the changing unit detects the small image for each small image. Coding is performed from the beginning of the GOP using the changed video coding parameter value, and the determination unit determines the small video based on the relation between the evaluation for each small video and the video coding parameter value. Each time, a video coding parameter value suitable for coding is determined.

  One aspect of the present invention is an acquisition step of acquiring different video coding parameter values used in coding each of a plurality of small video included in a large video, and the video coding parameter for each small video. A video code suitable for coding based on an evaluation step of performing an evaluation of coding performed on the small video using a value based on a predetermined index, and a relation between the evaluation and the video coding parameter value And a determination step of determining a quantization parameter value.

  One aspect of the present invention is a program that causes a computer to function as any one of the video coding parameter adjustment devices described above.

  According to the present invention, it is possible to adjust video coding parameters accurately and easily.

FIG. 1 is a functional block diagram showing a configuration of a video encoding apparatus according to a first embodiment of the present invention. It is a figure which shows the structure of the input imaging | video by the same embodiment. It is a flowchart which shows the encoding parameter automatic adjustment process by the embodiment. It is a figure which shows initialization of the GOP phase by the embodiment. It is a flowchart which shows the encoding parameter automatic adjustment process by 2nd Embodiment. It is a figure which shows calculation of (lambda) using the relationship of the generation code amount R and the encoding distortion D by the embodiment.

Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.
In this embodiment, a parameter adjustment function for adjusting a video coding parameter (also described as a coding parameter) is added to a video coding apparatus or a video coding system, and a test video is repeatedly input in a loop. The parameter adjustment function changes the coding parameter value in accordance with the cycle of the loop, and performs video coding and RD cost aggregation using the changed coding parameter value. Then, based on the relationship between the obtained coding parameter value and the RD cost, the optimum coding parameter value is automatically determined sequentially. Detailed embodiments will be described below.

First Embodiment
FIG. 1 is a functional block diagram showing the configuration of a video encoding device 101 according to the present embodiment. The video encoding device 101 is an example of a video encoding parameter adjustment device. In the same drawing, only functional blocks related to the present embodiment are extracted and shown. As shown in the figure, the video encoding device 101 includes a video input unit 102, a video encoding unit 103, a control unit 104, an RD cost calculation unit 105, an optimum parameter determination unit 106, and a memory 107.

  The video input unit 102 inputs a video signal. The video encoding unit 103 encodes a video signal. The control unit 104 changes the coding parameter value used when the video coding unit 103 performs coding. The RD cost calculation unit 105 calculates an RD cost of encoding by the video encoding unit 103. The optimal parameter determination unit 106 determines an optimal coding parameter value based on the relationship between the coding parameter value and the RD cost. The memory 107 stores various data. The memory 107 stores the coding parameter value and the RD cost calculated by the RD cost calculation unit 105 in association with each other.

  The video encoding device 101 has an operation mode for parameter adjustment, separately from the normal video encoding operation mode. In the normal video coding operation mode, the video coding 103 performs coding similar to that of the prior art on the input video, using a predetermined coding parameter value or the like. Below, the operation mode for parameter adjustment is explained.

  FIG. 2 is a diagram showing the configuration of an input video used for parameter adjustment. In the operation mode for parameter adjustment, an external device repeatedly reproduces a test video in a loop and inputs it to the video encoding device 101. The n-th loop (n is an integer of 1 or more) of the test video in the input video is described as loop #n. The test video is composed of videos of a plurality of scenes. The S (S is an integer of 1 or more) scenes included in the test video are sequentially described as scene # 0, scene # 1, ..., scene # (S-1). Each scene has a frame number longer than one GOP (Group of Picture). In addition, a black image for loop top detection is added to the end (or just before the start) of the test video of each loop.

FIG. 3 is a flow chart showing coding parameter adjustment processing according to the present embodiment.
The video input unit 102 takes the input video signal (input video) into the video encoding device 101 as image data. The video input unit 102 calculates an inter-frame difference value with the immediately preceding frame for each picture captured as image data, and outputs the calculated value to the control unit 104. The inter-frame difference value is a sum of absolute values of differences in luminance values at the same pixel position between the picture and the immediately preceding frame of the picture. Further, the video input unit 102 calculates an input video feature quantity such as an in-picture luminance value accumulated value which is an accumulation of luminance values for each picture, and outputs the calculated to the control unit 104. Also, the video input unit 102 outputs the input video to the video encoding unit 103. The video encoding unit 103 encodes an input video.

  The control unit 104 detects a scene change in the input video based on the inter-frame difference value received from the video input unit 102 and the amount of change thereof. Further, the control unit 104 identifies a frame that satisfies the scene change and the in-picture luminance value accumulated value is equal to or less than the threshold value as the first frame of the loop (step S101). The control unit 104 notifies the video encoding unit 103 when the head of the loop is detected.

  The control unit 104 starts adjusting the encoding parameters as follows when the head of the first loop is identified after the video encoding unit 103 starts video encoding of the input video.

  First, the control unit 104 sequentially changes the value of a specific coding parameter used by the video coding unit 103 during video coding for each head of the loop, and transmits the value to the video coding unit 103. For example, the cost cost_mode of the mode serving as an option at the time of mode determination of each CU (coding unit) is calculated by the following equation (1).

cost_mode = SATD_mode + λ_mode * R_mode (1)

  Here, an example will be described in which the optimum value of λ_mode used in the above equation (1) is determined. Note that R_mode is an estimated code amount of a mode serving as an option in each CU, and SATD_mode is a value obtained by performing Hadamard transform of the prediction error and summing.

  For λ_mode, different initial setting values λ′_mode [pt] [QP] are determined in advance according to the quantization parameter QP and the picture type pt. Therefore, the control unit 104 sets a quantization parameter QP to be processed (step S102). The control unit 104 sets λ_mode in Equation (1) when the quantization parameter QP and the picture type pt are the n-th loop (n is an integer of 1 or more and N or less and N is a predetermined upper limit). It is transmitted to the video encoding unit 103 to perform encoding using λ_mode [pt] [QP] calculated by the following equation (2) (step S103).

λ_mode [pt] [QP] = λ′_mode [pt] [QP] * (0.5 + 0.1 * n) (2)

  Therefore, λ_mode takes a different value for each picture type in the loop.

  The video encoding unit 103 encodes the test video of the n-th loop using λ_mode [pt] [QP] notified from the control unit 104 (step S104). At this time, the video encoding unit 103 performs encoding using a fixed quantization parameter set in advance, instead of the encoding of the normal fixed bit rate control performed by the control unit 104. For example, assuming that the quantization parameter in the I picture is QPi, the video encoding unit 103 fixedly adds +1 to +3 to the quantization parameter according to the picture type for the other picture types, and inputs Encode the video. For example, in the case of a GOP structure of SOP (Structure Of Pictures) size 8 (hereinafter, referred to as M = 8), the video encoding unit 103 QPi + 1, temporal_ID = for a B picture (hereinafter referred to as B0 picture) of temporal_ID = 0. The encoding is performed with QPi + 2 for the 1 and 2 B pictures (hereinafter, B1 and B2 pictures) and QPi + 3 for the temporal_ID = 3 B picture (hereinafter, B3 picture).

  Also, upon receiving from the control unit 104 the information indicating that the encoding target picture is the head of the loop, the video encoding unit 103 initializes the phase of the GOP. FIG. 4 is a diagram showing the initialization of the GOP phase. The video encoding unit 103 sets the beginning of the loop as an I picture at the beginning of the GOP, and encodes from this I picture. This makes it possible to encode the same picture every time with the same picture type, and to compare RD costs accurately. Even when the video encoding unit 103 receives, from the control unit 104, information indicating that the encoding target picture is the beginning of a scene, it initializes the phase of the GOP and sets the GOP beginning as an I picture.

  When the phase of the GOP can not be initialized due to the restriction of the function of the video encoding unit 103, the number of pictures in one loop of the test video is set to be an integral multiple of the number of pictures in GOP. Misalignment can be avoided.

  When encoding for one loop is performed in step S104 in FIG. 3, the video encoding unit 103 calculates the generated code amount R for each picture and the encoding distortion D, and transmits the amount to the RD cost calculation unit 105. The encoding distortion D is a sum of squares of pixel value differences of the original image and the decoded image. The control unit 104 outputs the picture type of each picture and the loop start flag when detecting the loop start to the RD cost calculation unit 105. The RD cost calculation unit 105 calculates the RD cost by the following equation (3) using the code amount R and the coding distortion D for each picture in the loop.

RDcost = D + λ * R (3)

  Note that different values of λ are determined in advance according to the quantization parameter QP and the picture type pt (see, for example, Non-Patent Document 1). The RD cost calculation unit 105 outputs the RD cost (RDcost) to the optimum parameter determination unit 106.

The optimum parameter determination unit 106 calculates the sum of RD costs (RDcost) for the same picture type pt in the loop #n, and stores it in the memory 107 as RDcost_Sum [n] [pt] (step S105). At this time, the optimum parameter determination unit 106 associates RDcost_Sum [n] [pt] with the used λ_mode [pt] [QP] and stores the associated value in the memory 107.
The video encoding apparatus 101 performs the above-described process for each of N loops (step S106).

  After the end of N times of loops, the optimum parameter determination unit 106 determines a loop number n_min that minimizes RDcost_Sum for each picture type based on RDcost_Sum [n] [pt] stored in the memory 107, The λ_mode at that time is set to λ_mode [n_min] [pt] (step S107). The optimal parameter determination unit 106 outputs λ_mode [n_min] [pt] for each picture type as an optimal parameter set.

  As described above, the set of λ_mode of each picture type in a certain quantization parameter value can be optimized. The video encoding device 101 performs the above-described processing by sequentially changing the value of the quantization parameter (step S108). This makes it possible to replace λ_mode from the initial set value λ′_mode [QP] [pt] to a value optimized by RD cost.

  It is assumed that the video encoding apparatus 101 is input with a video signal for performing a loop reproduction 21 times on a test video including seven scenes composed of 32 frames. The video encoding apparatus 101 performs the above processing and performs encoding while changing the encoding parameter little by little for each test video 1 loop. For example, the video encoding apparatus 101 encodes each of seven scenes using 21 different λ_modes for each loop with respect to quantization parameters of six different values and eight picture types, and sums RD costs. The coding parameters for which is the smallest are sequentially determined. In this case, the time required from the video input to the determination of the coding parameter is (number of QPs) 6 x (number of picture types) 8 x (number of loops) 21 x (number of scenes) 7 x (one frame of seconds) Number) 32/60 = 3763 seconds. As described above, it is possible to greatly streamline the parameter adjustment conventionally performed in operations such as video input, encoding, stream collection, acquisition of decoded video, aggregation and analysis of various logs, and a simulator.

Second Embodiment
In the present embodiment, another example of the automatic adjustment method of the coding parameter will be described. In the present embodiment, the video encoding apparatus 101 performs parameter adjustment in accordance with the reference order of pictures for each picture type, and further calculates the Lagrange coefficient λ for RD cost calculation for each scene. Therefore, more accurate parameter adjustment can be performed.

  FIG. 5 is a flowchart showing the encoding parameter automatic adjustment process according to the present embodiment. First, the video input unit 102 waits until the loop head of the video comes (step S201). In the meantime, the video encoding device 101 performs normal video encoding, and does not perform parameter adjustment processing. The top of the video loop is a scene change from the black image. The control unit 104 identifies the loop head from the value and the change amount of the inter-frame difference of the input video, and the luminance accumulation value of the video, and notifies the video encoding unit 103 of the loop top flag.

  The control unit 104 sets a new QP to be used for encoding in the video encoding unit 103 out of values (hereinafter referred to as QP) of quantization parameters not to be processed yet (step S202). In the present embodiment, in consideration of adjusting parameters having different values according to the QP, the QP to be adjusted is sequentially changed to 7, 12, 17, 22, 27, 32, 37, 42, 47. Although all the possible QPs may be adjusted, in practice, it is not problematic if parameter values for all QPs are obtained by interpolating parameter values obtained for discrete QPs. Also, since QP is generally +1 to +3 depending on the picture type on the basis of an I picture, this embodiment changes the QP and encodes in this way.

  The control unit 104 determines, from among the picture types that have not been selected as the process target, the picture types to be adjusted and counted from now (step S203). Here, the picture type is a picture having a similar reference relationship in the GOP. In the M = 8 GOP structure, the picture types are I, B0, B1, B2, B3 and so on. Bn represents a B picture of temporal_id = n.

  Subsequently, in steps S204 to S206, the video encoding device 101 calculates λ for RD cost calculation for each scene. First, the video encoding unit 103 encodes a test video for the first one loop included in the input video at QP-1. The RD cost calculation unit 105 calculates the generated code amount R and the encoding distortion D for each picture for the picture type to be aggregated or all the picture types, and calculates the accumulated value R [QP of the generated code amount R for each scene #s. -1] [s] and the accumulated value D [QP-1] [s] of the encoding distortion D are obtained (step S204). Here, s is a scene number.

  Subsequently, the video encoding apparatus 101 performs the same process on QP + 1 (step S205). That is, the video encoding unit 103 encodes the next one loop of test video included in the input video using QP + 1. The RD cost calculation unit 105 calculates the generated code amount R and the encoding distortion D for each picture for the picture type to be aggregated or all the picture types, and accumulates the generated value R of the generated code amount R for each scene #s. An accumulated value D [QP + 1] [s] of QP + 1] [s] and the encoding distortion D is obtained.

  FIG. 6 is a diagram showing the calculation of λ using the relationship between the generated code amount R and the encoding distortion D. Based on the RD characteristic curve shown in the figure, the RD cost calculation unit 105 calculates RD [.lambda.] For RD cost calculation for each scene #s, as the following equation (4).

λ [s] = (-1) * (R [QP + 1] [s] -R [QP-1] [s]) / (D [QP + 1] [s] -D [QP-1] [s]) ... (4)

  In FIG. 5, the control unit 104 changes the value of the coding parameter used when coding the test video of the next loop, and outputs the value to the video coding unit 103 (step S207). For example, at the time of motion search, the cost cost_me of each search point of each PU (Prediction Unit) is calculated by the following equation (5).

cost_me = SAD + λ_me * R (5)

  An example will be described in which the optimum value of λ_me used in the above equation (5) is determined as a coding parameter. Here, R is an estimated amount of code necessary to encode the MV (motion vector) of the search point, and SAD is the sum of absolute values of prediction errors.

  In general, λ_me has different initial setting values for each picture type and QP. Therefore, in the loop #n (n is an integer of 1 or more and N or less, N is a predetermined upper limit value), the control unit 104 calculates λ_me in the equation (5) by the following equation (6) It will be a value.

λ_me [n] = λ_me * (0.5 + 0.1 * n) (6)

  The video encoding unit 103 encodes a test image for one loop, and the RD cost calculation unit 105 calculates the generated code amount R and the encoding distortion D for each picture for the picture type of the aggregation target, and the one loop An accumulated value R [n] of the generated code amount R for a minute and an accumulated value D [n] of the encoding distortion D are obtained. The RD cost calculation unit 105 obtains the RD cost for the scene #s of loop #n according to the following equation (7) (step S208).

cost [s] [n] = D [n] + λ [s] * R [n] ... (7)

  The optimum parameter determination unit 106 associates the picture type of the aggregation target, the RD cost cost [s] [n] calculated by the RD cost calculation unit 105, and λ_me [n] with each other and writes the result in the memory 107.

  The control unit 104 determines whether the number of loops has reached N (step S209). If the control unit 104 determines that the number of loops is less than N, the video encoding device 101 returns again to step S207, changes the encoding parameter for the next loop, and encodes one loop. Perform and aggregate RD costs.

  If the control unit 104 determines that the number of loops has reached N (step S 209), the optimum parameter determination unit 106 determines, for each scene, the coding parameter with the smallest RD cost for the type of picture to be aggregated. (Step S210). Specifically, the optimum parameter determination unit 106 refers to the memory 107, and the loop number n_min [s] in which the RD cost cost [s] [n] calculated for each scene #s for the picture type of the aggregation target is minimized. To identify. The optimum parameter determination unit 106 acquires, from the memory 107, the parameter λ_me [n_min [s]] which is λ_me [n] associated with the specified loop number n_min [s].

  Further, in the present embodiment, the optimum parameter determination unit 106 takes the average of λ_me obtained for all scenes for the picture types to be aggregated, and outputs the calculated average as the QP and the adjustment value of λ_me for the picture type. Also, this value is used in coding in the subsequent parameter adjustment of the picture type. Thereby, the reference image can be adjusted with the adjusted value, and adjustment with higher accuracy is possible.

  In another embodiment, the optimum parameter determination unit 106 determines λ_me [n_min [s]] of each scene for the picture type to be aggregated, and then adjusts them for the QP and the picture type to adjust λ_me of each scene. While outputting as a value, it can be used in encoding of each scene at the time of parameter adjustment of the picture type after that. By this, it is possible to perform parameter adjustment for each scene in substantially the same procedure.

  The control unit 104 determines whether the process has been completed for all picture types (step S211). If the control unit 104 determines that there is an unprocessed picture type (step S211: NO), the control unit 104 returns to step S203, selects a new picture type, and starts parameter adjustment. If the control unit 104 determines that the process has been completed for all picture types (step S211: YES), the control unit 104 performs the determination of step S212.

  The control unit 104 determines whether the process has ended for all the QPs (step S212). If the control unit 104 determines that there is an unprocessed QP (step S212: NO), the control unit 104 returns to step S202 and starts parameter adjustment using a new QP. If the control unit 104 determines that the process has ended for all the QPs (step S212: YES), the entire process illustrated in FIG. 5 ends.

  Although the example in which the parameter adjustment function is added to the video encoding device 101 has been described above, the parameter adjustment function may be added to a video encoding system configured with a plurality of devices.

  According to the embodiment described above, the video coding parameter adjustment device includes an acquisition unit, an evaluation unit, and a determination unit. For example, the video coding parameter adjustment device is the video coding device 101, the acquisition unit is the control unit 104, the evaluation unit is the RD cost calculation unit 105, and the determination unit is the optimum parameter determination unit. The acquisition unit acquires different video coding parameter values used in coding each of the plurality of small videos included in the large video. For example, the large video is an input video of the video encoding device 101, and the small video is one or more scenes included in the test video to be loop-reproduced. The evaluation unit performs, for each small video, the evaluation of coding performed on the small video using the video coding parameter value based on a predetermined index. The predetermined index is, for example, an RD cost. The determination unit determines a video coding parameter value suitable for coding based on the relation between the evaluation and the video coding parameter value.

  The video coding parameter adjustment device may further include a detection unit, a change unit, and a coding unit. For example, the detection unit and the change unit are the control unit 104, and the encoding unit is the video encoding unit 103. The detection unit detects the head of the loop based on the video image feature amount of the large video. The change unit changes a video coding parameter value used when coding a small video at the beginning of the loop detected by the detection unit. The encoding unit initializes the GOP phase at the beginning of the loop detected by the detecting unit, and encodes the small image from the beginning of the GOP using the video encoding parameter value changed by the changing unit for each loop.

  Also, the changing unit sequentially changes the target picture type in accordance with the order of the reference relationship in the GOP. The evaluation unit may evaluate, for each picture type, the coding performed on the picture included in the small video using the video coding parameter value based on a predetermined index. The determination unit determines a video coding parameter value suitable for coding for each of the picture types, based on the relationship between the evaluation for each of the picture types and the video coding parameter value.

  If the predetermined index is an RD cost obtained by linearly combining the generated code amount R and the coding degradation D, the evaluation unit determines the quantization parameter value of the adjustment target in the small images of the two loops included in the large image. The slope of the RD characteristic curve is determined based on the generated code amount R and the encoding degradation D when encoding is performed using quantization parameter values before and after QP), and the RD cost is calculated based on the determined slope. You may determine the Lagrange coefficient used for calculation.

  In addition, the detection unit may detect the head of each of the plurality of small videos included in each loop based on the video feature amount of the large video. The encoding unit initializes the GOP phase at the beginning of the small image detected by the detection unit, and performs encoding from the beginning of the GOP using the video encoding parameter value changed by the changing unit for each small image. The determination unit determines a video coding parameter value suitable for coding for each small video based on the relation between the evaluation for each small video and the video coding parameter value.

  According to the above-described embodiment, it is possible to perform optimization and adjustment of video coding parameters more efficiently and accurately than the conventional method. Further, according to the above-described embodiment, identification of the loop head and scene head of the test video, initialization of the GOP phase according to it, aggregation of RD cost for each picture type and each scene, Lagrange coefficient used for RD cost calculation The adjustment of the video coding parameter can be performed with high accuracy by the automatic calculation of.

  The above-described video encoding device 101 includes a CPU (Central Processing Unit), a memory, an auxiliary storage device, and the like connected by a bus, and by executing a video encoding parameter adjustment program, a video input unit 102 and a video encoding unit The control unit 104 functions as an apparatus including the control unit 104, the RD cost calculation unit 105, and the optimum parameter determination unit 106. Note that all or part of the functions of the video input unit 102, the video encoding unit 103, the control unit 104, the RD cost calculation unit 105, and the optimum parameter determination unit 106 can be implemented using an application specific integrated circuit (ASIC) or programmable logic (PLD). It may be realized using hardware such as Device) or FPGA (Field Programmable Gate Array). The video coding parameter adjustment program may be recorded on a computer readable recording medium. The computer readable recording medium is, for example, a portable medium such as a flexible disk, a magneto-optical disk, a ROM, a CD-ROM, or a storage device such as a hard disk built in a computer system. The video coding parameter adjustment program may be transmitted via a telecommunication line.

  The embodiment of the present invention has been described in detail with reference to the drawings. However, the specific configuration is not limited to this embodiment, and includes design and the like within the scope of the present invention.

DESCRIPTION OF SYMBOLS 101 ... Video coding apparatus, 102 ... Video input part, 103 ... Video coding part, 104 ... Control part, 105 ... RD cost calculation part, 106 ... Optimal parameter determination part, 107 ... Memory

Claims (7)

An acquisition unit for acquiring different video coding parameter values used in coding each of a plurality of small videos included in the large video;
An evaluation unit which performs, based on a predetermined index, an evaluation of coding performed on the small video using the video coding parameter value for each small video;
A determination unit that determines a video coding parameter value suitable for coding based on the relation between the evaluation and the video coding parameter value;
Video coding parameter adjustment device comprising: The large image includes the small image to be loop-played,
A detection unit that detects the head of the loop based on the video image feature amount of the large video;
A changing unit that changes the video coding parameter value used when coding the small video at the head of the loop detected by the detection unit;
The GOP (Group of Picture) phase is initialized at the head of the loop detected by the detection unit, and the small video is recorded from the head of the GOP using the video coding parameter value changed by the changing unit for each loop. And an encoding unit for encoding
The video coding parameter adjustment device according to claim 1. The evaluation unit performs, based on a predetermined index, evaluation of coding performed on a picture included in the small video using the video coding parameter value for each picture type.
The determination unit determines a video coding parameter value suitable for coding for each of the picture types, based on a relation between the evaluation for each of the picture types and the video coding parameter value.
The video coding parameter adjustment device according to claim 1. The large image includes the small image to be loop-played,
The predetermined index is an RD cost obtained by linearly combining the generated code amount R and the coding degradation D,
The evaluation unit generates a generated code amount R when encoding is performed on the small image of the two loops included in the large image using quantization parameter values before and after the quantization parameter value to be adjusted. And determining the slope of the RD characteristic curve based on the coding deterioration D, and determining the Lagrange coefficient used to calculate the RD cost based on the determined slope.
The video coding parameter adjustment device according to claim 1. The large image includes a plurality of small images to be loop-played,
The detection unit detects the beginning of the small video based on the video feature amount of the large video.
The encoding unit initializes a GOP (Group of Picture) phase at the beginning of the small image detected by the detection unit, and uses the image encoding parameter value changed by the changing unit for each small image. Encoding from the beginning of the GOP,
The determination unit determines a video coding parameter value suitable for coding for each small video, based on the relation between the evaluation for each small video and the video coding parameter value.
The video coding parameter adjustment device according to claim 2. Obtaining different video coding parameter values used in coding each of a plurality of small videos included in the large video;
An evaluation step of performing, based on a predetermined index, evaluation of encoding performed on the small video using the video coding parameter value for each small video;
Determining a video coding parameter value suitable for coding based on the relation between the evaluation and the video coding parameter value;
And a video coding parameter adjustment method.   A program that causes a computer to function as the video coding parameter adjustment device according to any one of claims 1 to 5.

Images