RU2839915C1 - Method and device for predictive video coding - Google Patents

Abstract

FIELD: physics.

SUBSTANCE: invention relates to video coding technologies. During interframe prediction in the process of making decision on the optimal mode for the current prediction unit, information on the executed mode is determined based on the executed mode, wherein information on performed mode includes optimal mode for step and costs for optimal mode for step; and, based on the information on the executed mode, determining whether to skip the intra-frame prediction mode when making a decision.

EFFECT: high encoding efficiency.

10 cl, 6 dwg

Description

[1] This application claims priority from Chinese Patent Application No. 202011461361.8, filed on December 11, 2020, which is incorporated herein by reference in its entirety.

Field of technology to which the present invention relates

[2] The present disclosure relates to the field of video coding technologies, for example, relates to a method and device for predictive video coding.

Prior art of the present invention

[3] High Efficiency Video Coding (HEVC) is the latest generation video coding standard, and it doubles the compression efficiency of H.264. To improve the compression efficiency, HEVC uses many sophisticated technologies. According to one feature, HEVC uses a quadtree partitioning method in which each block is called a coding unit (CU) and has a maximum size of 64 x 64, and is recursively divided into four coded sub-blocks, and the smallest CU has a size of 8 x 8, and therefore the block segmentation in HEVC is more flexible and has more block types than H.264. According to another feature, HEVC uses prediction modes for each CU, and the prediction mode corresponds to the prediction unit (PU). There are two types of prediction modes: inter-frame prediction modes (including Merge&Skip_2N x 2N, Inter_2N x 2N, Inter_2N x N, Inter_N x 2N, Inter_2N x nU, Inter_2N x nD, Inter_nL x 2N, Inter_nR x 2N, InterN x N and the like) and intra-frame prediction modes (including Intra_2N x 2N and Intra_N x N). I-frames are used only for the intra-frame prediction mode, and P-frames and B-frames are used in all prediction modes. In the HEVC standard, the coding cost is calculated for each mode and the prediction mode with the minimum coding cost is selected, which achieves higher data compression efficiency.

[4] For P-frames and B-frames of the HEVC standard, the intra-frame prediction mode is also an important prediction mode. In the case of motion sequence, since the motion is accompanied by the appearance of a new scene and the disappearance of an old scene, the best matching blocks cannot be found for some coding units in each frame. For each coding unit, the prediction performance of the intra-frame prediction mode is undoubtedly higher than that of the inter-frame prediction mode. However, in general, the best matching blocks are found for most of the coding units in neighboring frames in the time domain, and therefore the selection rate of intra-frame prediction in P-frames and B-frames is low.

[5] Compared with the H.264 standard, the intra-frame prediction mode in the HEVC standard is greatly improved. The HEVC standard supports 32 directional prediction modes and two non-directional prediction modes (flat mode and DC mode), while the H.264 standard only supports eight directional prediction modes. Therefore, the complexity of intra-frame prediction in the HEVC standard is greater than that of the H.264 standard, and it is necessary to perform the calculation of the cost to achieve the desired rate and distortion (RD cost) in each prediction mode. The calculation process for determining the mode is very complex and time-consuming. Therefore, when unnecessary mode selection for rate and distortion optimization (RDO) is reduced, the resource cost of the encoder is reduced greatly, so that the coding rate increases and the coding cost decreases.

Brief summary of the present invention

[6] According to embodiments of the present invention, a method and apparatus for predictive video coding are provided that eliminate the disadvantages of high coding resource consumption, low coding efficiency, and high coding costs caused by performing an intra-frame prediction mode in some inter-frame coding methods used in practice.

[7] According to a first feature of embodiments of the present invention, a method for predictive video coding is provided, comprising the following steps:

[8] determining, according to the executed mode, information about the executed mode in the process of deciding on the best mode for the current prediction unit in the inter-picture prediction, wherein the information about the executed mode includes a temporary best mode and a cost for the temporary best mode; and

[9] determining, based on information about the mode being executed, whether to skip the intra-frame prediction mode in the decision-making process.

[10] According to a second aspect of embodiments of the present disclosure, a predictive video coding apparatus is provided, which includes:

[11] an execution mode information determining module configured to determine, according to an execution mode, information about the execution mode in a process of deciding a best mode for a current prediction unit in inter-frame prediction, wherein the execution mode information includes a temporary best mode and a cost for the temporary best mode; and

[12] an intra-frame prediction mode skip determination module configured to determine, based on information about the mode being executed, whether to skip the intra-frame prediction mode in the decision-making process.

[13] According to a third feature of the embodiments of the present disclosure, an electronic device is further provided that includes a memory, a processor, and a computer program stored in the memory and executed on the processor, wherein the processor, upon loading and executing the program, is caused to perform the method described above.

[14] According to a fourth feature of the embodiments of the present disclosure, a computer-readable storage medium is further provided, on which a computer program is stored, wherein the computer program, when loaded into a processor and executed by it, causes the processor to perform the method described above.

Brief description of the figures

[15] Fig. 1 shows a flow chart of an algorithm of a method for predictive video coding according to some embodiments of the present disclosure;

[16] Fig. 2 shows a simplified flow chart of a process for making a decision on a prediction unit PU of an inter-frame coding mode according to some embodiments of the present disclosure;

[17] Fig. 3 shows a flow chart of an algorithm for skipping an intra-frame mode decision according to some embodiments of the present disclosure;

[18] Fig. 4 shows a flow chart of an algorithm for skipping an intra-frame mode decision according to some embodiments of the present disclosure;

[19] Fig. 5 shows a block diagram of an apparatus for predictive video coding according to some embodiments of the present disclosure; and

[20] Fig. 6 shows a simplified block diagram of an electronic device according to some embodiments of the present disclosure;

Detailed disclosure of the present invention

[21] The present disclosure will be described in detail hereinafter with reference to the corresponding figures in conjunction with embodiments. It should be understood that the specific embodiments described herein are used only to explain the present disclosure and are not intended to limit the present disclosure. In addition, it should also be noted that for simplicity of description, only some portions are shown in the figures, rather than all complete structures associated with the present disclosure.

[22] Fig. 1 shows a flow chart of a method for predictive video coding according to some embodiments of the present disclosure. The embodiments are applicable to a predictive coding step in video coding. Predictive coding means predicting a value of a current sample based on one or more values of encoded samples using a model or method, coding a difference value between the actual value and the predicted value of the sample, and transforming, quantizing and entropy coding the predicted residual rather than the original pixel value, whereby coding efficiency is significantly improved. Predictive coding includes an intra-frame prediction mode (Intra) and an inter-frame prediction mode (Inter). I-frames are used only for the intra-frame prediction mode Intra, and P-frames and B-frames are used in all prediction modes.

[23] For the P-frames and B-frames of the HEVC standard, the intra-frame prediction mode is also an important prediction mode. The intra-frame mode mainly means predicting the value of the current pixel based on the coded pixels and the spatial correlation of the pixels in the video image, so that the spatial redundancy of the video image is removed. In the intra-frame coding of the luminance component of the HEVC standard, five types of PU prediction units are used: 4 × 4, 8 × 8, 16 × 16, 32 × 32 and 64 × 64, for each PU size, there are 35 prediction modes: DC, flat and 33 angular modes, for the chrominance component, there are five prediction modes corresponding to DC, flat, horizontal, perpendicular, and a prediction mode corresponding to the luminance component. For each prediction mode, it is necessary to perform calculations on the cost of obtaining the rate and distortion, and the process of such calculations for deciding the mode is complex and time-consuming. In the case where unnecessary mode selection for rate and distortion optimization (RDO) is reduced by using a fast intra-frame PU coding algorithm, the resource cost of the encoder is reduced significantly, so that the coding rate increases and the coding cost decreases.

[24] Taking this into account, according to embodiments, the intra-frame mode is processed quickly in inter-frame coding. The compression speed of the encoder is improved by simplifying the complex procedure of deciding on the intra-frame mode in inter-frame coding in the encoder. By improving the compression speed of the encoder, the performance of the transcoding service is improved, so that server resources are saved and the cost of transcoding is reduced.

[25] It should be noted that this method according to the embodiments is applicable to any hybrid coding architecture using blocks. This method is applicable to all encoders that meet the requirements of the HEVC standard, and other encoders that comply with standards such as AVS1, H.264, VP8, VP9, AVS2, AVS3, AV1 and VVC can directly use this method after adjustment. The encoder according to the embodiments is applicable to both the transcoding service and the real-time coding service of the mobile communication terminal.

[26] According to an embodiment of the method, the following stages are provided.

[27] In step S110, according to the mode being executed, information about the mode being executed is determined in the process of deciding the best mode for the current prediction unit in the inter-frame prediction, wherein the information about the mode being executed includes a temporary best mode and a cost for the temporary best mode.

[28] At step S110, based on the information about the mode being executed, it is determined whether to skip the intra-frame prediction mode in the decision-making process.

[29] Fig. 2 shows a simplified flow chart of a process for deciding on a prediction unit PU of an interframe coding mode according to some embodiments of the present disclosure. In the process of deciding on the interframe mode of the PU, the following processes are provided.

[30] First of all, the MERGE/SKIP mode of 2N x 2N is executed. If the recursion condition is met, the recursive mode is executed for the coding unit CU, after the recursion is executed, the CU with the size of the current block is returned, and the intra-frame prediction mode (Intra) is continuously executed. After the intra-frame prediction mode (Intra) is completed, the inter-frame mode of 2N x 2N is executed (the best matching block is selected by forward and backward search), and then the bidirectional search (BIDIR) of 2N x 2N is executed (only bidirectional search is performed to obtain the best matching block). The 2N x N and N x 2N types of modes are continuously executed. A mode decision is made after the completion of one mode to obtain a temporary best mode (tempBestMode). Here, N is a positive integer.

[31] After completing the N x 2N mode, in the case where the temporary best mode tempBestMode=2N x N, the modes 2N x nU and 2N x nD are continuously executed, and a comparison is made to obtain the final best mode BestMode; in the case where the temporary best mode tempBestMode=N x 2N, the modes nL x 2N and nR x 2N are continuously executed, and a comparison is made to obtain the final best mode (BestMode); and in the case where the temporary best mode tempBestMode !=2N x N && tempBestMode !=N x 2N, 2N x nU, the modes 2N x nD, nL x 2N, and nR x 2N are continuously executed, and a comparison is made to obtain the final best mode (BestMode).

[32] It should be noted that in the case where recursion for a CU is performed on the current block, the cost of the best mode BestMode is compared with the cost of the recursion to determine whether to perform recursive partitioning of the CU being coded.

[33] Furthermore, for modes such as 2N x N, N x 2N, 2N x nU, 2N x nD, nL x 2N, and nR x 2N that involve two PUs, MERGE/SKIP, forward and backward search, and bidirectional search must be performed for each PU to obtain the best-matching block.

[34] According to some embodiments, for the above-mentioned decision-making process, in the case where a decision on the best mode for the inter-frame prediction mode is performed on the current prediction unit PU, the corresponding information on the mode to be executed is obtained based on the mode to be executed, and a decision on whether to skip the intra-frame prediction (Intra) mode is made in the above-mentioned decision-making process based on the information on the mode to be executed.

[35] As an example, the information about the mode to be executed includes a temporary best mode and a cost for the temporary best mode. According to embodiments, various methods are provided for determining whether to skip the intra-frame mode Intra based on various information about the mode to be executed.

[36] In the example (illustrative) implementation scenario, skipping the Intra intra-frame mode means skipping a part of the processes of the Intra mode. Thus, when executing the Intra intra-frame mode, the Intra intra-frame decision is changed so that the RDO optimization is not performed in the Intra mode, so as to greatly reduce the complexity of the Intra intra-frame mode and reduce the coding time. The fast intra-frame decision is immediately performed based on the inter-frame coding information of the PU for the current PU, to determine whether to skip the Intra mode without additional computation and thus achieve high cost efficiency.

[37] According to some embodiments, the costs of the temporary best mode include the costs of the SATD for the temporary best mode. As shown in Fig. 3, step S120 includes the following substeps.

[38] In sub-stage S11, the SATD cost for the best intra-frame prediction mode is determined by a coarse decision process of the SATD in the intra-frame prediction mode for a prediction unit.

[39] According to some embodiments, when performing an intra-frame prediction mode, from among all available intra-frame prediction modes, the prediction mode with the lowest sum of absolute transformed differences (SATD) cost, together with the low-complexity SATD cost, is selected and determined to be the best intra-frame prediction mode.

[40] In sub-stage S12, according to the prediction unit size and together with the SATD cost in the temporal best mode and the SATD cost in the best intra-frame prediction mode, it is determined whether to skip the subsequent process of the intra-frame prediction mode.

[41] In this process, after obtaining the SATD cost in the temporal best mode InterSATDCost and the SATD cost in the intra-frame best prediction mode RModeSATDCost, it is determined whether to skip the subsequent process of the intra-frame mode according to the prediction units PUs with different sizes.

[42] PU prediction units with different sizes include 2N x 2N PU units and N x N PU units. 2N x 2N PU units refer to PU units with sub-blocks of size 64 x 64, 32 x 32, 16 x 16, and 8 x 8, and N x N PU units refer to PU units with sub-blocks of size 4 x 4.

[43] According to some embodiments, in the case where the size of the current prediction unit PU is N x N, sub-stage S12 provides the following lower-level sub-stages.

[44] At sub-stage S121, it is determined whether the skip transform mode will be applied for the first prediction unit with size 4x4.

[45] According to embodiments, in the case where the size of the current prediction unit PU is N x N, N x N is a specific intra-frame coding mode in which intra-frame coding is performed only on coding units of PUs that can be continuously divided into four 4 x 4 in divided 8 x 8 PUs. For coding a 4 x 4 PU, the specific coding mode is a transform skip (TSKIP) mode.

[46] The transform skip mode is a technology introduced in the HEVC standard, using a plurality of intra-frame or inter-frame 4 x 4 blocks. It has an obvious effect on slowing down the video sequence when the scene changes, so that the compression efficiency is improved. Since the transform skip modes of the four obtained 4 x 4 PUs are connected, the transform skip modes of the first 4 x 4 PUs are determined according to the embodiments, so that the fact that a transform unit (TU) corresponds to the first 4 x 4 PU means making a decision on the transform skip mode, so that it is determined whether to skip the unexecuted subsequent process in the intra-frame mode.

[47] In sub-stage S122, based on the result of determining the cost of SATD in the temporary best mode, the cost threshold of the mode is determined.

[48] According to some embodiments, in the case where the skip transform mode is applied for the first 4 x 4 PU, the cost threshold of the mode is determined as the product of the cost of the SATD in the temporary best mode and a first tuning factor. In the case where the skip transform mode is not applied for the first 4 x 4 PU, the cost threshold of the mode is determined as the product of the cost of the SATD in the temporary best mode and a second tuning factor. The second tuning factor is less than the first tuning factor. Thus, the cost threshold of the mode is determined according to the following formula:

SATDCostThreshold=firstTskip ? InterSATDCost * al: InterSATDCost* b1.

[49] Here, SATDCostThreshold denotes the mode cost threshold, firstTskip? denotes whether the first 4 x 4 PU is in the skip transform mode (i.e., whether the first 4 x 4 PU is a TSKTP block), InterSATDCost denotes the SATD cost in the temporary best mode, al denotes the first tuning multiplier, and b1 denotes the second tuning multiplier.

[50] In case firstTskip is 1 and the first 4 x 4 PU is a TSKIP block, SATDCostThreshold=InterSATDCost * a1. In case firstTskip is 0 and the first 4 x 4 PU is not a TSKIP block, SATDCostThreshold=InterSATDCost * b1.

[51] It should be noted that since TSKIP has a positive effect on coding efficiency, in the case where the first 4 x 4 PU is a TSKIP block, there is a high probability that the subsequent blocks are TSKIP blocks. Therefore, the mode cost threshold is improved to eliminate skipping. Therefore, the value of al is adjusted to a value greater than b1. The specific values of a1 and b1 are adjusted according to actual requirements. For example, a1 is 0.5, and b1 is 0.3.

[52] In sub-step S123, the subsequent process of the intra-frame prediction mode is skipped after determining that the SATD cost in the best intra-frame prediction mode is greater than the mode cost threshold.

[53] After the mode cost threshold is determined, the SATD cost in the best intra-frame prediction mode is compared with the mode cost threshold. When the SATD cost in the best intra-frame prediction mode is greater than the mode cost threshold, the intra-frame coding cost is high. The intra-frame mode consumes coding rate, and therefore the subsequent intra-frame process is skipped.

[54] According to some embodiments, the determination condition is further adjusted in conjunction with the adjustment factors. Various requirements for speeding up the operation are taken into account by adjusting the adjustment factors. For example, the decision to skip a subsequent process in the intra-frame mode is determined by the following conditional expression:

[55] Thus, in the case where RModeSATDCost * c is greater than d * SATDCostThreshold, the coding cost of performing intra-frame mode is high, and therefore the subsequent process in intra-frame mode is skipped. The end values are tuning multipliers, and they are adjusted according to the actual requirements. For example, both end values are equal to 1.

[56] According to some embodiments, in the case where the size of the current prediction unit PU is 2N x 2N, the decision to skip the subsequent process in the intra-frame mode is determined together with the recursive encoding condition. As shown in FIG. 2, in the case where the executed mode includes a recursive mode of the coding unit CU, the information about the executed mode further includes a value of the ratio at which the coded sub-block corresponding to the current CU is an intra-frame prediction sub-block. The intra-frame prediction sub-block is the sub-block in which the determined temporal best mode is an intra-frame prediction mode. For example, for four sub-blocks of CUs obtained by dividing the current CU, in the case where the number of sub-blocks (i.e., intra-frame sub-blocks) in which the predicted best mode is an intra-frame mode is one of the four sub-blocks of CUs, the value of the ratio is 1/4. In the case where the number of sub-blocks in which the predicted best mode is an intra-frame mode is equal to two out of four CU sub-blocks, the value of the ratio is 2/4.

[57] According to some embodiments, in the case where the size of the current prediction unit PU is 2N x 2N, sub-stage S12 provides the following lower-level sub-stages.

[58] The subsequent process of the intra-frame prediction mode is skipped after determining that the SATD cost in the best intra-frame prediction mode is greater than the SATD cost in the temporal best mode, and the ratio value is less than a predetermined ratio threshold.

[59] For example, the decision to skip a subsequent process in intraframe mode is determined using the following conditional expression:

[60] Here, Ra denotes a ratio value at which the CU sub-block corresponding to the current CU is an intra-frame sub-block, and the values of e, f, and g respectively denote tuning factors, and they are adjusted according to actual requirements. For example, the values of e, f, and g are respectively adjusted to 1, 1, and 0.75.

[61] With respect to the above formula, in the case where RModeSATDCost * e>f * InterSATDCost, the cost of intra-frame prediction coding is high. In the case where the Ra value of the CU sub-block is small, for example, less than g, the probability that the best mode for the corresponding CU is the intra-frame mode is low. Therefore, the subsequent process of the intra-frame prediction mode is immediately skipped to prematurely terminate the execution of the intra-frame mode, so that the prediction time is saved and the coding efficiency is improved.

[62] According to some embodiments, in the case where it is determined in steps S11 and S12 that the subsequent process of the intra-frame prediction mode is not skipped, the subsequent process of the intra-frame prediction mode continues to be continuously executed. For example, a candidate for the prediction mode roughly selected by the SATD and the most possible mode (MOP) form a set of candidates for the prediction mode. The MOP is calculated based on the intra-frame prediction modes of neighboring PUs, and the neighboring PUs are the upper block, the left block, and the upper left block, respectively. Then, RDO optimization is performed on the candidates for the prediction mode to obtain the best intra-frame prediction mode.

[63] In another example (illustrative) implementation scenario, skipping the intra-frame mode means skipping the entire process of the intra-frame mode, and the intra-frame mode is not performed at all to save encoding time. In the process of deciding the best inter-frame mode, the decision to skip the intra-frame mode is made based on the information of the temporary best mode, and the temporary best mode is a variable depending on various conditions, and it is characterized by adaptive adjustment and high stability.

[64] According to some embodiments, the costs of the temporary best mode include the costs of the RDO for the temporary best mode. As shown in Fig. 4, step S120 includes the following substeps.

[65] In sub-stage S21, it is determined whether the executed mode includes a recursive mode of the coding unit.

[66] At sub-step S22, based on the results of considering the temporal best mode and/or the RDO costs for the temporal best mode, it is determined whether to skip the intra-frame prediction mode in the decision making process.

[67] According to some embodiments, in the case where the executed mode includes a recursive mode of the coding unit CU, the information about the executed mode further includes one or a combination of the following elements: a sum of the RDO costs of the coded sub-blocks corresponding to the current coding unit, a first flag with information indicating whether the current coding unit is an all-zero block, and a second flag with information indicating whether the coded sub-block corresponding to the current coding unit includes an intra-frame prediction sub-block. The intra-frame prediction sub-block is the sub-block in which the predicted best mode is an intra-frame prediction mode. Sub-step S22 provides the following lower-level sub-step:

[68] skipping the intraframe mode in the decision-making process after determining that one or more of the following conditions are met.

[69] Condition 1: The sum of the RDO costs is greater than the RDO costs for the temporary best mode.

[70] Since the modes executed are modes executed with RDO, after the modes are executed, the corresponding RDO information (i.e. information about the executed mode) appears.

[71] After performing the recursive mode of the CU unit, the RDO cost for each sub-block of the CU unit is determined, and the sum of the RDO costs for all the sub-blocks of the CU is calculated to obtain the sum of the RDO costs (i.e., the recursion cost). In the case where the sum of the RDO costs is greater than the RDO cost of the temporal best mode, the recursion cost is greater than the cost of the temporal best mode, and therefore the recursive division of the CU unit is not performed. Therefore, the coding rate is saved. In the intra-frame mode, the coding rate is consumed, and therefore the coding rate is saved by skipping the intra-frame mode in the case where the sum of the RDO costs is greater than the RDO cost of the temporal best mode.

[72] Condition 2: The first flag information indicates that the current coding unit is a block of all zeros.

[73] According to this condition, in the case where the first flag information indicates that the current block is an all-zero block, because the coding rate consumed by the all-zero block is low, and the intra-frame mode consumes the coding rate, and these two statements are contradictory, and therefore the probability that the best mode for the all-zero block is the intra-frame mode is low, and the intra-frame mode is skipped in the case where the current block is an all-zero block.

[74] Condition 3: The second flag information indicates that the coded sub-block corresponding to the current coding unit does not include an intra-frame prediction sub-block.

[75] According to this condition, in the case where the sub-block being coded does not include an intra-frame prediction sub-block (i.e., the sub-block being coded does not include a sub-block for which the predicted best mode is an intra-frame mode), the probability that the best mode for the current block is an intra-frame mode is low. Therefore, in the case where the sub-block does not include an intra-frame sub-block, the intra-frame mode is skipped immediately.

[76] According to some embodiments, in order to improve the determination accuracy and reduce losses, the above three conditions are combined for consideration. For example, conditions 1 and 2 are combined. In the case where conditions 1 and 2 are met simultaneously, the intra-frame mode is skipped, as shown in the following defining conditional expression:

[77] here denotes the RDO cost for a sub-block of a CU unit, and denotes the sum of RDO costs for sub-blocks, BestModeRdCost is the RDO cost for the temporary best mode, BestMode->CBF denotes the first flag information used to indicate that the current block being encoded is an all-zeros block, splitIntra denotes the second flag information used to indicate that the sub-blocks include an intra-frame block, a2 and b2 are tuning factors that are adjusted according to actual requirements. For example, a2 and b2 are configured as 1 by default.

[78] According to the above defining conditional expression, in the case where the conditions and !BestMode->CBF are executed simultaneously, i.e., the current block is an all-zero block, and the sum of the RDO costs is greater than the RDO cost of the temporary best mode, the intra-frame mode is skipped. Alternatively, in the case where the sub-blocks do not include the intra-frame sub-block, the intra-frame mode is also skipped.

[79] According to some embodiments, in the case where the executed mode does not include the recursive mode of the CU unit, sub-stage S22 provides for the next lower-level sub-stage.

[80] In the case where the temporal best mode is the SKIP mode, the intra-frame prediction mode is skipped.

[81] According to embodiments, since the SKIP mode saves coding rate and the intra-frame mode consumes coding rate, in order to save coding rate, the intra-frame mode is skipped immediately in the case where the temporary best mode is determined to be the SKIP mode based on the mode being executed.

[82] According to some embodiments, in the case where the intra-frame mode cannot be skipped during the execution of sub-steps S21 and S22, the embodiments further provide the following lower-level sub-step.

[83] The SATD cost in the 2N x 2N intra-frame mode is determined by performing 2N x 2N intra-frame coding for the current coding unit, and the N x N intra-frame prediction mode is skipped after determining that the SATD cost in the 2N x 2N intra-frame mode is greater than the SATD cost in the temporal best mode.

[84] According to embodiments, in the case where the intra-frame mode cannot be skipped during the execution of sub-steps S21 and S22, for the current coding unit CU, 2N x 2N intra-frame coding is first performed and the SATD cost in the 2N x 2N intra-frame mode is determined. Then, the SATD cost in the 2N x 2N intra-frame mode is compared with the SATD cost in the temporal best mode to determine whether to skip the N x N intra-frame mode.

[85] For example, the decision to skip an N x N intraframe mode is determined using the following conditional expression:

[86] Here, Intra2N x 2NSATDCost denotes the SATD cost in the intra-frame 2N x 2N mode, BestModeSATDCost denotes the SATD cost in the temporal best mode, and the values of a3 and b3 denote tuning factors, and they are tuned according to the actual requirements. For example, the values of a3 and b3 are tuned as 10 and 11, respectively.

[87] According to embodiments, information about the mode to be executed is obtained based on the mode to be executed when deciding on the best mode for the current prediction unit in the inter-prediction mode. The information about the mode to be executed includes a temporal best mode and a cost for the temporal best mode. Then, based on the information about the mode to be executed, a decision is made whether to skip the intra-prediction mode in the decision process, so that the use of the intra-prediction mode is reduced, the resource cost of the encoder is reduced significantly, the encoding speed is improved, and the encoding cost is reduced. According to embodiments, by deciding to skip the intra-prediction mode, the encoding speed is increased by 15% under the condition that the compression ratio degradation in the encoder is 0.2%. In the context of applying online transcoding, the server computing resource is saved by 15% with a slight loss in image quality, so that the transcoding cost is reduced.

[88] Fig. 5 shows a block diagram of an apparatus for predictive video coding according to some embodiments of the present disclosure. The apparatus includes the following modules:

[89] an execution mode information determining module 510 configured to determine, according to an execution mode, information about the execution mode in a process of deciding a best mode for a current prediction unit in inter-picture prediction, wherein the execution mode information includes a temporary best mode and a cost for the temporary best mode; and

[90] an intra-frame prediction mode skip determination module 520 configured to determine, based on information about the mode being executed, whether to skip the intra-frame prediction mode in the decision-making process.

[91] According to some embodiments, the cost of the temporal best mode includes the cost of the SATD for the temporal best mode. The intra-frame prediction mode skip determination module 520 includes the following submodules:

[92] a coarse SATD decision submodule configured to determine the SATD cost for the best intra-frame prediction mode using a coarse SATD decision process in the intra-frame prediction mode for a prediction unit; and

[93] a first skip determination submodule in an intra-frame mode, configured to determine whether to skip a subsequent process of an intra-frame prediction mode according to a prediction unit size and in conjunction with the SATD cost in the temporal best mode and the SATD cost in the best intra-frame prediction mode.

[94] According to some embodiments, the prediction unit size is N x N, where N is a positive integer. The intra-frame first gap determination submodule is configured to:

[95] determine whether the first 4 x 4 prediction unit can be given a skip transform mode;

[96] determine the cost threshold for the mode based on the result of determining the SATD costs in the temporary best mode; and

[97] skip the subsequent process of the intra-frame prediction mode after determining that the SATD cost of the best intra-frame prediction mode is greater than the mode cost threshold.

[98] According to some embodiments, the intra-frame first skip determination submodule is configured to:

[99] define the mode cost threshold as the product of the SATD cost in the temporal best mode and the first tuning factor after determining that the skip transform mode is applied for the first 4 x 4 prediction unit; and

[100] define the mode cost threshold as the product of the SATD cost in the temporal best mode and the second tuning factor after determining that for the first 4 x 4 prediction unit the skip transform mode is not applied, and the second tuning factor is less than the first tuning factor.

[101] According to some embodiments, in a case where the prediction unit size is 2N x 2N, the executed mode includes a recursive mode of the coding unit, and the information about the executed mode includes a value of a ratio at which a coded sub-block corresponding to the current coding unit is an intra-frame prediction sub-block; wherein the intra-frame prediction sub-block is a sub-block in which the predicted best mode is an intra-frame prediction mode; and

[102] the submodule for determining the first gap in the intra-frame mode is configured to:

[103] skip the subsequent process of the intra-frame prediction mode after determining that the SATD cost in the best intra-frame prediction mode is greater than the SATD cost in the temporal best mode, and the ratio value is less than a predetermined ratio threshold.

[104] According to some embodiments, the cost of the temporal best mode includes the cost of the RDO in the temporal best mode. The intra-frame prediction mode skip determination module 520 includes the following submodules:

[105] a recursion determination submodule configured to determine whether the executed mode includes a recursive mode of the coding unit; and

[106] a second intra-frame mode skip determination submodule configured to determine whether to skip the intra-frame prediction mode in the decision-making process, the determination being made based on the results of determining at least one of the following values: a temporal best mode and an RDO cost for the temporal best mode.

[107] According to some embodiments, after determining that the mode to be performed includes a recursive mode of the coding unit, the information about the mode to be performed further includes one or a combination of the following information: a sum of the RDO costs of the coded sub-blocks corresponding to the current coding unit, a first flag with information indicating whether the current coding unit is an all-zero block, and a second flag with information indicating whether the coded sub-block corresponding to the current coding unit includes an intra-frame prediction sub-block, wherein the intra-frame prediction sub-block is a sub-block in which the predicted best mode is an intra-frame prediction mode, and

[108] the submodule for determining the second gap in the intra-frame mode is configured to:

[109] skip the intraframe mode in the decision-making process after determining that one or more of the following conditions are met:

[110] the sum of the RDO costs is greater than the RDO costs for the temporary best mode,

[111] The first flag information indicates that the current coding unit is a block of all zeros, and

[112] the second flag information indicates that the coded sub-block corresponding to the current coding unit does not include an intra-frame prediction sub-block.

[113] According to some embodiments, after determining that the mode being executed does not include a recursive mode of the coding unit, the second skip determination submodule in the intra-frame mode is configured to:

[114] skip the intra-frame prediction mode after determining that the temporal best mode is the SKIP mode.

[115] According to some embodiments, the costs of the temporary best mode further include the costs of the SATD for the temporary best mode. The device further includes:

[116] a 2N x 2N intra-frame coding module configured to determine a SATD cost for the 2N x 2N intra-frame coding by performing the 2N x 2N intra-frame coding for the current coding unit based on the result of not skipping the intra-frame prediction mode in the decision process; and

[117] An N x N intra-frame mode skip determination module configured to skip the N x N intra-frame prediction mode after determining that the SATD cost in the 2N x 2N intra-frame mode is greater than the SATD cost in the temporal best mode.

[118] It should be noted that the above-mentioned apparatus for predictive video encoding according to the embodiments of the present disclosure performs the methods for predictive video encoding according to the embodiments of the present disclosure, and is equipped with functional modules for obtaining advantages corresponding to the methods performed.

[119] Fig. 6 shows a simplified block diagram of an electronic device according to some embodiments of the present disclosure. As shown in Fig. 6, the electronic device includes a processor 610, a memory 620, an input hardware 630, and an output hardware 640. The number of processors 610 in the electronic device is one or more, and Fig. 6 shows one processor 610 as an example. The processor 610, memory 620, input hardware 630, and output hardware 640 in the electronic device are connected by a bus or other means, and Fig. 6 shows a bus connection as an example.

[120] Memory 620, as a computer-readable storage medium, is configured to store software programs, a computer-executable program and a module, such as program instructions or a module, corresponding to the methods according to embodiments of the present disclosure. Processor 610 executes various functional application programs and data processing in the electronic device, that is, executes the above-mentioned methods, it does this by executing the software program, instructions and module stored in memory 620.

[121] The memory 620 mainly contains a program storage area and a data storage area.

[122] The program storage area stores an operating system and an application program that is required to perform at least one function; and the data storage area stores data created during use of the terminal. Moreover, the memory 620 includes a high-speed random access memory device, and further comprises a non-volatile memory, such as at least one hard disk drive, flash memory or other non-volatile semiconductor memory. According to some embodiments, the memory 620 includes memory devices located remotely relative to the processor 610, and these remote memory devices are connected to the electronic device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile network and a combination thereof.

[123] Input apparatus 630 receives input digital or text information and creates a key input signal related to a user configuring or operating an electronic device. Output apparatus 640 includes a display device, such as a display screen.

[124] According to some embodiments of the present disclosure, a storage medium is further provided that contains computer-executable instructions. The computer-executable instructions, when executed by a processor on a server, cause the processor to implement a method according to any of the embodiments of the present disclosure.

[125] Based on the embodiments described above, those skilled in the art can clearly understand that the present disclosure can be implemented either by software and a necessary general hardware platform, or by hardware. Based on such understanding, the technical solutions of the present disclosure, which make a significant or partial contribution to the known technical solutions, can be implemented in the form of a software product. The software product for a computer can be stored on a computer-readable storage medium, such as a floppy disk, a read-only memory (ROM), a random access memory (RAM), a flash memory, a hard disk, or an optical disk of a computer, and it contains several instructions, so that a computing device (a personal computer, a server, or a network device) executes the methods according to a plurality of embodiments of the present disclosure. The computer-readable storage medium is a non-volatile computer-readable storage medium.

[126] It should be noted that, according to the above-mentioned embodiments of the device, the plurality of its blocks and modules are divided only according to functional logic, but they are not limited to the above-mentioned division, provided that the corresponding functions are implemented. In addition, the specific names of the plurality of functional blocks are indicated only for the purpose of distinguishing them from each other, and they are not intended to limit the scope of legal protection of the present disclosure.

Claims (36)

1. A method for predictive video coding, comprising the following stages: determining, according to an executable mode, information about the executable mode in a process of deciding on the best mode for the current prediction unit in inter-frame prediction, wherein the information about the executable mode includes a temporal best mode and a cost for the temporal best mode; wherein the cost for the temporal best mode includes a sum of absolute transformed differences (SATD) for the temporal best mode; determining the SATD cost for the best intra-frame prediction mode using a coarse decision process for the SATD in the intra-frame prediction mode for a prediction unit; and determining whether a skip transform mode can be applied to the first prediction unit of size 4 x 4 after determining that the prediction unit size is N x N; determining a mode cost threshold based on a result of determining the SATD cost in the temporal best mode; and skipping a subsequent process of the intra-frame prediction mode after determining that the SATD cost in the best intra-frame prediction mode is greater than the mode cost threshold; or in a case where the prediction unit size is 2N x 2N, the executed mode includes a recursive mode of a coding unit, and the information about the executed mode includes a ratio value at which a coded sub-block corresponding to the current coding unit is an intra-frame prediction sub-block; skipping a subsequent process of the intra-frame prediction mode after determining that the SATD cost in the best intra-frame prediction mode is greater than the SATD cost in the temporal best mode, and the ratio value is less than a predetermined ratio threshold; where N is a positive integer, the intra-frame prediction sub-block is the sub-block in which the predicted best mode is an intra-frame prediction mode. 2. The method according to paragraph 1, characterized in that the determination of the threshold of costs for the mode based on the result of determining the costs for SATD in the temporary best mode includes the following stages: determining the mode cost threshold as the product of the SATD cost in the temporal best mode and the first tuning factor after determining that the skip transform mode is applied for the first 4 x 4 prediction unit; and determining the mode cost threshold as the product of the SATD cost in the temporary best mode and a second tuning factor after determining that the first 4 x 4 prediction unit does not apply the skip transform mode, and the second tuning factor is less than the first tuning factor. 3. The method according to item 1, characterized in that the cost of the temporary best mode includes the cost of rate and distortion optimization (RDO) for the temporary best mode; and the method further comprises the steps of: determining whether the mode being executed includes the recursive mode of the coding unit after determining that the cost of the temporary best mode includes the cost of the RDO for the temporary best mode; and determining, based on the results of determining at least one of the following quantities: a temporal best mode and an RDO cost for the temporal best mode, whether to skip the intra-frame prediction mode in the decision making process. 4. The method according to paragraph 3, characterized in that after determining that the mode to be executed includes a recursive mode of the coding unit, the information about the mode to be executed further includes at least one of the following elements: a sum of the RDO costs of the coded sub-blocks corresponding to the current coding unit, a first flag with information indicating whether the current coding unit is an all-zero block, and a second flag with information indicating whether the coded sub-block corresponding to the current coding unit includes an intra-frame prediction sub-block, wherein the intra-frame prediction sub-block is a sub-block in which the predicted best mode is an intra-frame prediction mode, and determining, based on the results of determining at least one of the following values: the temporal best mode and the RDO costs for the temporal best mode, whether to skip the intra-frame prediction mode in the decision-making process, includes the stage of: skipping the intra-frame prediction mode in the decision process after determining that at least one of the following conditions is met: the sum of the RDO costs is greater than the RDO cost for the temporal best mode, the first flag information indicates that the current coding unit is an all-zero block, and the second flag information indicates that the coded sub-block corresponding to the current coding unit does not include an intra-frame prediction sub-block. 5. The method according to claim 3, characterized in that after determining that the mode being executed does not contain a recursive mode of the coding unit, determining, based on the results of determining at least one of the following values: a temporary best mode and the RDO costs for the temporary best mode, whether to skip the intra-frame prediction mode in the decision-making process, includes the stage of: skipping the intra-frame prediction mode after determining that the temporal best mode is the SKIP mode. 6. The method according to any of paragraphs 3-5, characterized in that the temporary best mode cost additionally includes the cost of the SATD for the temporary best mode; and after determining, based on the information about the mode being executed, whether the intra-frame prediction mode should be skipped in the decision-making process, the method further provides the steps of: obtaining the SATD cost for 2N x 2N intra-frame coding by performing 2N x 2N intra-frame coding for the current coding unit based on the result of not missing the intra-frame prediction mode in the decision process; and skipping the N x N intra-frame prediction mode after determining that the SATD cost in the 2N x 2N intra-frame mode is greater than the SATD cost in the temporal best mode. 7. A device for predictive video coding, comprising: an executable mode information determining module configured to determine, according to an executable mode, executable mode information in a process of deciding a best mode for a current prediction unit in inter-picture prediction, wherein the executable mode information includes a temporal best mode and a cost for the temporal best mode; and the cost for the temporal best mode includes a sum of absolute transformed differences (SATD) for the temporal best mode; and a coarse SATD decision submodule configured to determine a SATD cost for the best intra-frame prediction mode using a coarse SATD decision process in the intra-frame prediction mode for a prediction unit; and a submodule for determining the first gap in intraframe mode, configured to: determining whether a transform skip mode can be applied to a first prediction unit of size 4 x 4 after determining that the prediction unit size is N x N; determining a mode cost threshold based on a result of determining the SATD cost in the temporal best mode; and skipping a subsequent process of the intra-frame prediction mode after determining that the SATD cost in the best intra-frame prediction mode is greater than the mode cost threshold; or an intra-frame prediction mode skip determination module configured to determine, based on information about the executed mode, whether to skip the intra-frame prediction mode in the decision process, wherein the skipping of the intra-frame prediction mode includes a partial skipping of the intra-frame prediction mode and a complete skipping of the intra-frame prediction mode; skipping the subsequent process of the intra-frame prediction mode in a case where the prediction unit size is 2N x 2N, the executed mode includes a recursive mode of the coding unit, and the information about the executed mode includes a ratio value at which the coded sub-block corresponding to the current coding unit is an intra-frame prediction sub-block after determining that the SATD cost in the best intra-frame prediction mode is greater than the SATD cost in the temporal best mode, and the ratio value is less than a predetermined ratio threshold; where N is a positive integer, the intra-frame prediction sub-block is the sub-block in which the predicted best mode is an intra-frame prediction mode. 8. The device according to item 7, in which the submodule for determining the first gap in the intra-frame mode is additionally configured with the ability to: determining the mode cost threshold as the product of the SATD cost in the temporal best mode and the first tuning factor after determining that the skip transform mode is used for the first 4 x 4 prediction unit; and determining the mode cost threshold as the product of the SATD cost in the temporal best mode and a second tuning factor after determining that the first 4 x 4 prediction unit does not apply the skip transform mode, and the second tuning factor is less than the first tuning factor. 9. An electronic device for predictive video coding, comprising: a memory, a processor, and a computer program stored in the memory and executed on the processor, wherein the processor, after loading and executing the program, is caused to perform the method that is defined in any of paragraphs 1-6. 10. A non-volatile computer-readable data carrier in which a computer program is stored, wherein the computer program, when loaded and executed by a processor, causes the processor to perform the method defined in any of paragraphs 1-6.