Video coding method and system, medium, equipment and program product

Abstract

The invention provides a video coding method and system, a medium, equipment and a program product, and relates to the technical field of video processing, and the method comprises the steps: determining an agent task; the proxy task is used for representing the performance quality state of the candidate hyper-parameter combination; defining a genetic algorithm solution space and a fitness function; the genetic algorithm solution space is used for defining a search range; in a genetic algorithm solution space, calling a fitness function to map individuals into numerical scores on the basis of a result of an agent task; the individuals are hyper-parameter combinations; enabling the selected individuals to enter a mating pool to form a parent population according to numerical scores and a set selection strategy; performing population operation on parent population circulation according to a set probability until a set iteration step number is met or the proxy task meets a preset threshold value, and outputting a hyper-parameter of an entropy coding skipping mechanism; and performing video coding based on the hyper-parameter of the entropy coding skipping mechanism. According to the invention, dual benefits of compression efficiency and calculation overhead can be realized.

Description

Technical Field

[0001] This application relates to the field of video processing technology, and in particular to a video encoding method, system, medium, device and program product. Background Technology

[0002] In the continuous iteration of end-to-end video coding models, EEM's entropy coding skip mode still uses a manually set set of hyperparameters. As the model structure, weights, and training objectives are updated, these prior parameters gradually become mismatched with the current network, causing performance to degrade from optimal to suboptimal. Simultaneously, the sensitivity of end-to-end models to the skip mechanism varies significantly at different bitrates, while the current framework forces the use of the same set of hyperparameters, failing to adapt to different bitrates. Furthermore, manual parameter tuning relies on expert experience and repeated trial and error, which is time-consuming and labor-intensive, severely restricting the efficiency of algorithm iteration. Therefore, those skilled in the art urgently need to address the problems of mismatch between manually set parameters and the model, performance imbalances across bitrates, and high iteration costs. Summary of the Invention

[0003] The purpose of this application is to provide a video coding method, system, computer-readable storage medium, electronic device, and computer program product that can iteratively output optimal hyperparameters without manually finding parameter sets.

[0004] To address the aforementioned technical problems, this application provides a video encoding method, the specific technical solution of which is as follows:

[0005] Determine the proxy task; the proxy task is used to characterize the performance of candidate hyperparameter combinations.

[0006] Define the solution space and fitness function of the genetic algorithm; the solution space of the genetic algorithm is used to define the search range.

[0007] In the solution space of the genetic algorithm, the fitness function is invoked to map individuals to numerical scores based on the results of the proxy task; the individuals are combinations of hyperparameters.

[0008] Based on the numerical scores and the set selection strategy, the selected individuals will be placed into the mating pool to form the parent population;

[0009] The parent population is subjected to population operations in a loop with a set probability until a set number of iterations is met or the proxy task meets a preset threshold. The hyperparameters of the entropy encoding skip mechanism are then output.

[0010] Video encoding is performed based on the hyperparameters of the entropy coding skipping mechanism.

[0011] Optionally, determining the agent task includes:

[0012] Determine the type of proxy task; the proxy task type includes zero sample, few sample, and many sample.

[0013] The number of sequence runs corresponding to the hyperparameter combination is determined based on the agent task type.

[0014] Optionally, the process of defining the solution space of the genetic algorithm includes:

[0015] Determine the key parameters corresponding to chromosomes in the genetic algorithm; the key parameters include the skip function, the basic skip probability constant, and the frame sequence scaling function.

[0016] Several specific chromosomes are determined based on the key parameters; the solution space of the genetic algorithm consists of all the specific chromosomes.

[0017] Optionally, after defining the solution space and fitness function of the genetic algorithm, the following also applies:

[0018] The skip rate calculation formula for entropy coding is defined based on the key parameters; the skip rate calculation formula is used to indicate whether to skip the current frame, and the skip rate is negatively correlated with the bit rate and positively correlated with the distortion.

[0019] Optionally, before selecting individuals to enter the mating pool to form the parent population based on the numerical scores and the set selection strategy, the process further includes:

[0020] The set selection strategy is determined from a selection strategy pool, which includes at least roulette selection strategies and tournament selection strategies.

[0021] Optionally, before performing population operations on the parent population cycle with a set probability, the method further includes:

[0022] Determine the type of population operation; the type of population operation includes heredity, mutation, crossover, and replication;

[0023] Determine the population operation probability corresponding to each of the aforementioned population operation types.

[0024] This application also provides a video encoding system, including:

[0025] The task determination module is used to determine the proxy task; the proxy task is used to characterize the performance of candidate hyperparameter combinations.

[0026] The genetic algorithm definition module is used to define the genetic algorithm solution space and fitness function; the genetic algorithm solution space is used to define the search range.

[0027] The individual scoring module is used to map individuals to numerical scores based on the results of the proxy task by calling the fitness function in the solution space of the genetic algorithm; the individual is a combination of hyperparameters.

[0028] The parent population generation module is used to select individuals into the mating pool to form the parent population based on the numerical score and the set selection strategy.

[0029] The iteration module is used to perform population operations on the parent population cyclically with a set probability until a set number of iteration steps is met or the proxy task meets a preset threshold, and output the hyperparameters of the entropy encoding skip mechanism.

[0030] The video encoding module is used to perform video encoding based on the hyperparameters of the entropy encoding skipping mechanism.

[0031] This application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the video encoding method described above.

[0032] This application also provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the video encoding method described above when it calls the computer program in the memory.

[0033] This application also provides a computer program product, including a computer program that, when executed, implements the steps of the video encoding method described above.

[0034] This application provides a video coding method, comprising: determining a proxy task; the proxy task being used to characterize the performance of candidate hyperparameter combinations; defining a genetic algorithm solution space and a fitness function; the genetic algorithm solution space being used to define a search range; under the genetic algorithm solution space, calling the fitness function to map individuals to numerical scores based on the results of the proxy task; the individuals being hyperparameter combinations; entering the selected individuals into a mating pool to form a parent population according to the numerical scores and a set selection strategy; performing population operations on the parent population cyclically with a set probability until a set number of iterations is met or the proxy task meets a preset threshold, and outputting hyperparameters of an entropy coding skipping mechanism; and performing video coding based on the hyperparameters of the entropy coding skipping mechanism.

[0035] This application defines the hyperparameter solution space of the entropy coding skipping mechanism, encodes the key parameters in the video coding skipping mechanism as chromosomes of a genetic algorithm, and searches for the optimal combination of entropy coding skipping mechanism parameters through the evolutionary process of the genetic algorithm, thereby performing video coding. By using proxy tasks, the high evaluation cost that originally required complete encoding-decoding-rate-distortion calculations is significantly reduced, significantly shortening the parameter tuning time. Leveraging the global search and local optima-escaping capabilities of the genetic algorithm, it can still approach the global optimum with a high probability in a discrete, non-convex hyperparameter space. Furthermore, each proxy task can be distributed across multiple cores or clusters, improving computational efficiency. By adopting the skipping mechanism hyperparameters, they can be directly reused in actual coding, resulting in a significant reduction in video bitrate with almost no loss of visual quality, bringing dual benefits in compression efficiency and computational cost reduction.

[0036] This application also provides a video encoding system, a computer-readable storage medium, an electronic device, and a computer program product, which have the above-mentioned beneficial effects, and will not be repeated here. Attached Figure Description

[0037] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of this application. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0038] Figure 1 A flowchart illustrating a video encoding method provided in an embodiment of this application;

[0039] Figure 2 This is a schematic diagram of the video encoding process provided in an embodiment of this application;

[0040] Figure 3 This is a schematic diagram of a video encoding system structure provided in an embodiment of this application;

[0041] Figure 4 This is a structural diagram of an electronic device provided in an embodiment of this application. Detailed Implementation

[0042] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0043] The End-to-End Intelligent Coding (EEM) direction within the AVS (Audio & Video Coding Standards) working group. Existing EEM frameworks, in the entropy coding process for motion and residual information, involve an entropy coding skip mode, where predicted values ​​are directly transmitted without residual coding. The function for deciding the skip mode is:

[0044] ;

[0045] Where C is a manually designed constant value for different frames (such as 0.16 and 0.23). It also involves manually designed functions and related hyperparameters.

[0046] As EEM continues to iterate, some proposals have suggested manually changing the parameter set corresponding to the skip mode. This is because, during model iteration, manually designed parameter sets gradually become incompatible with the model, becoming suboptimal options. Furthermore, end-to-end video coding models at different bitrates require significantly different parameter sets, but the current entropy coding skip mechanism uses the same set of hyperparameters because manually finding a suitable parameter set is extremely time-consuming and labor-intensive.

[0047] See Figure 1 , Figure 1 The flowchart illustrates a video encoding method provided in this application embodiment. To solve the above-mentioned technical problems, this application provides a video encoding method, the specific technical solution of which is as follows:

[0048] S101: Determine the proxy task; the proxy task is used to characterize the performance of candidate hyperparameter combinations.

[0049] S102: Define the solution space and fitness function of the genetic algorithm; the solution space of the genetic algorithm is used to define the search range;

[0050] S103: In the solution space of the genetic algorithm, the fitness function is called to map individuals to numerical scores based on the results of the agent task; the individuals are combinations of hyperparameters.

[0051] S104: Based on the numerical scores and the set selection strategy, the selected individuals will be placed into the mating pool to form the parent population;

[0052] S105: Perform population operations on the parent population in a cycle with a set probability until the set number of iterations is met or the proxy task meets the preset threshold, and output the hyperparameters of the entropy encoding skip mechanism.

[0053] S106: Perform video encoding based on the hyperparameters of the entropy coding skipping mechanism.

[0054] In step S101, the aim is to determine the proxy task. During this process, the proxy task type can be determined. The proxy task type includes zero samples, few samples, and many samples, thereby determining the number of sequence runs corresponding to the hyperparameter combination based on the proxy task type.

[0055] Taking multiple sample types as an example, when determining the proxy task, a small number of video sequences can be selected to calculate the video quality of the end-to-end video codec model under the current entropy coding skipping mechanism. (Distortion, such as PSNR, SSIM, VMAF, etc.) and compression efficiency (Rate, bitrate).

[0056] Calculate the score of individual P in the agent task. The higher the score, the better:

[0057] .

[0058] In step S102, the solution space of the genetic algorithm is first defined. In one feasible implementation, key parameters corresponding to chromosomes in the genetic algorithm can be determined. These key parameters include a skip function, a basic skip probability constant, and a frame sequence scaling function. Then, several specific chromosomes are determined based on these key parameters. At this point, the solution space of the genetic algorithm consists of all of these specific chromosomes.

[0059] Specifically, the solution space can be split / encoded to obtain key parameters (corresponding to chromosomes in genetic algorithms), which may include, but are not limited to, skip functions. The basic skip probability constant C and the frame sequence scaling function .

[0060] Under the above definition, the skip rate calculation formula for entropy coding can be further defined based on key parameters. The skip rate calculation formula for entropy coding is as follows:

[0061] ;

[0062] The skip rate calculation formula is used to indicate whether to skip the current frame, and the skip rate is negatively correlated with the bit rate and positively correlated with the distortion.

[0063] Skip functions are used to determine the differences in skip probabilities between multiple frames, such as the one currently used in EEM. This application defines the superspace for skipped functions, including but not limited to the following types:

[0064] ;

[0065] ;

[0066] ;

[0067] ;

[0068] ;

[0069] in To define the parameters of the hyperspace, for Different values ​​correspond to different hyperspaces. It is possible to simultaneously... By considering multiple possible values, the larger the hyperspace becomes, and the greater the probability of finding the global optimum. However, the search cost also increases.

[0070] The base skip probability constant C: The base probability used to determine the skip probability, including but not limited to the following values:

[0071] ;

[0072] Hyperspace of parameters such as frame sequence scaling function ,in This indicates the relative position of the current frame in the sequence, taking into account... For the case where the value is [0,1], The definition includes, but is not limited to, the following types:

[0073] ;

[0074] ;

[0075] ;

[0076] ;

[0077] ;

[0078] In simple terms, the solution space determines which "recipes" can be used to select ingredients; the fitness function is a "score sheet" for measuring the quality of a recipe; and the "skip rate calculation formula" is just a "key ingredient" in the recipe, whose value directly affects the final score, thus determining the life or death of a certain recipe (chromosome) in the genetic algorithm.

[0079] The solution space of a genetic algorithm is composed of chromosomes. Once the skip function, hyperspace parameters, and basic skip probability constants are determined, a specific chromosome can be obtained. All possible chromosomes together constitute the solution space of the genetic algorithm. The larger the range, i.e., the larger the hyperspace parameters... The more values ​​of and the closer the step size of the basic skip probability constant C, the larger the solution space, the higher the probability of finding the global optimum, and the higher the corresponding search cost.

[0080] In step S103, the fitness function needs to be called to map the individual to a numerical score based on the result of the agent task. The higher the score, the better. The scoring formula is as follows:

[0081] ;

[0082] Subsequently, individuals selected based on the numerical scores and the established selection strategy are placed into a mating pool to form the parent population. Prior to this, the established selection strategy can be determined from a selection strategy pool, which includes at least roulette wheel selection and tournament selection strategies. By dynamically switching between different established selection strategies, premature convergence or getting trapped in local optima is avoided, maintaining population diversity. Simultaneously, the online configurability of the strategy allows the algorithm to respond instantly to changes in problem characteristics, automatically balancing exploration and exploitation without manual intervention. This design significantly improves the quality of the global search, accelerates convergence, and reduces parameter tuning costs.

[0083] The tournament selection strategy assigns a probability to each individual that is selected that is proportional to its fitness (the higher the fitness, the greater the probability of selection). Similar to roulette, the selection area is allocated proportionally.

[0084] Tournament selection strategies, such as the most direct approach of obtaining the highest-scoring TopK.

[0085] The selected individuals will enter the mating pool to form the parent population FP. At this point, the population operation type can be determined, and then the population operation probability corresponding to each of the aforementioned population operation types can be determined. Population operation types include heredity, mutation, crossover, and replication. By implementing heredity, mutation, crossover, and replication on the parent population with predetermined probabilities, new allele combinations can be continuously injected at the chromosome level, thereby expanding the coverage of the search space. The probabilistic operation mechanism not only preserves the genetic information of high-fitness individuals, but also effectively avoids premature convergence through random perturbation, taking into account the dual needs of global exploration and local refinement. The synergistic effect of crossover and replication promotes the rapid diffusion of excellent solutions, while appropriate mutation continuously replenishes diversity, enabling the population to maintain robust evolutionary momentum during the iteration process, ultimately significantly improving the convergence speed and optimal solution quality of the algorithm.

[0086] No specific limits are set for the probabilities here. For example, the probabilities of heredity, mutation, crossover, and replication can be {0.4, 0.1, 0.4, 0.1}, respectively.

[0087] Repeat the above steps until the iteration count is completed, or the proxy task has reached the pre-set threshold.

[0088] This application's embodiments significantly reduce the high evaluation cost, which originally required complete encoding-decoding-rate-distortion calculations, by using proxy tasks, thus significantly shortening the parameter tuning time. Leveraging the genetic algorithm's ability to perform global search and escape local optima, it can still approximate the global optimum with a high probability in discrete, non-convex hyperparameter spaces. Each proxy task can be distributed across multiple cores or clusters, improving computational efficiency. By employing a skipping mechanism, hyperparameters can be directly reused in actual encoding, significantly reducing the video bitrate while maintaining almost no loss in visual quality, resulting in a dual benefit of reduced compression efficiency and computational overhead.

[0089] As can be seen from the above embodiments, this application achieves a balanced optimization of video compression efficiency and computational complexity by combining a genetic algorithm with an adaptive entropy coding skipping mechanism. By applying this application, it is possible to intelligently decide whether to skip the entropy coding stage based on the characteristics of the video content, thereby significantly reducing computational resource consumption while maintaining good compression quality. The key lies in utilizing the global search capability of the genetic algorithm to automatically search for decision parameters for entropy coding skipping, enabling it to adapt to end-to-end video codec models under different bitrates and versions, rather than using fixed thresholds or conditional judgments.

[0090] In the practical application of this application, the key parameters in the video coding skip mechanism (such as the skip function selection, the basic skip probability constant C, and the frame sequence scaling function) can be defined by defining the hyperparameter solution space of the entropy coding skip mechanism. The first step involves encoding the genetic algorithm's chromosomes. Secondly, a proxy task is designed, comprehensively considering factors such as video quality (e.g., PSNR, SSIM) and compression efficiency (bitrate) to guide the evolutionary direction of the genetic algorithm. Through the evolutionary process of the genetic algorithm (including inheritance, mutation, crossover, and replication operations), the optimal combination of entropy coding skipping mechanism parameters is searched for. Finally, the optimized parameters are integrated into the end-to-end video coding framework to construct an adaptive decision-making module.

[0091] See Figure 2 , Figure 2 This is a schematic diagram of the video encoding process provided in the embodiments of this application. The following is a further explanation of the specific process of video encoding based on the hyperparameters of the entropy coding skipping mechanism:

[0092] Entropy coding, as the final compression step in end-to-end video coding technology, directly impacts the overall coding performance. Figure 2 This paper demonstrates a video encoding and decoding process based on inter-frame prediction and feature reconstruction, which is mainly divided into two parts: I-frame processing and P-frame processing.

[0093] When processing I-frames:

[0094] Input original image: Original image Input into the system.

[0095] I-frame Encoder: Original Image The image is encoded by an I-frame encoder to generate a compressed representation.

[0096] I-frame Decoder: The compressed representation is decoded by the I-frame decoder to obtain the reconstructed image. It serves as a reference frame for subsequent P-frame processing.

[0097] When processing P-frames:

[0098] Current frame Reconstructed image from the previous frame The features of the current frame are obtained by the feature extraction module respectively. Features of the previous frame Using motion estimation models to... and Perform motion estimation and generate motion information. Sports Information Through motion encoder and motion entropy encoding model ( Figure 2 The motion entropy estimation model in the model is encoded, and then compressed into a binary bitstream using an arithmetic coding algorithm.

[0099] The compressed binary bitstream is decoded using a motion entropy coding model (the motion entropy estimation model in the figure) and an arithmetic coding algorithm, and then decoded again using a motion decoder to obtain the reconstructed motion information. Utilizing reconstructed motion information Features of the previous frame Perform alignment prediction to generate aligned features. .

[0100] In the residual coding and decoding stage, the features of the current frame are... and aligned features The concat input is fed into the RC encoder (Residual Coding Encoder) for encoding. It is then compressed into a binary bitstream using the residual entropy coding model (the residual entropy estimation model in the figure) and an arithmetic coding algorithm.

[0101] The compressed binary bitstream is decoded using a residual entropy coding model (the residual entropy estimation model in the diagram) and an arithmetic coding algorithm, and then decoded again using an RC decoder to obtain the reconstructed residual. .

[0102] Finally, perform confidence-based feature reconstruction: utilizing the reconstructed residuals. and aligned features By combining a confidence-based feature reconstruction model, the final reconstructed features are generated. The final reconstructed features The image reconstruction module generates a reconstructed image. .

[0103] See Figure 3 , Figure 3 This is a schematic diagram of a video encoding system structure provided in an embodiment of this application. The system includes:

[0104] The task determination module is used to determine the proxy task; the proxy task is used to characterize the performance of candidate hyperparameter combinations.

[0105] The genetic algorithm definition module is used to define the genetic algorithm solution space and fitness function; the genetic algorithm solution space is used to define the search range.

[0106] The individual scoring module is used to map individuals to numerical scores based on the results of the proxy task by calling the fitness function in the solution space of the genetic algorithm; the individual is a combination of hyperparameters.

[0107] The parent population generation module is used to select individuals into the mating pool to form the parent population based on the numerical score and the set selection strategy.

[0108] The iteration module is used to perform population operations on the parent population cyclically with a set probability until a set number of iteration steps is met or the proxy task meets a preset threshold, and output the hyperparameters of the entropy encoding skip mechanism.

[0109] The video encoding module is used to perform video encoding based on the hyperparameters of the entropy encoding skipping mechanism.

[0110] Based on the above embodiments, as a preferred embodiment, the task determination module includes:

[0111] A type determination unit is used to determine the type of proxy task; the proxy task type includes zero samples, few samples, and many samples.

[0112] The sequence determination unit is used to determine the number of sequence runs corresponding to the hyperparameter combination based on the agent task type.

[0113] Based on the above embodiments, as a preferred embodiment, the process of defining the solution space of the genetic algorithm includes:

[0114] Determine the key parameters corresponding to chromosomes in the genetic algorithm; the key parameters include the skip function, the basic skip probability constant, and the frame sequence scaling function.

[0115] Several specific chromosomes are determined based on the key parameters; the solution space of the genetic algorithm consists of all the specific chromosomes.

[0116] Based on the above embodiments, as a preferred embodiment, it further includes:

[0117] The calculation formula definition module is used to define the skip rate calculation formula for entropy coding based on the key parameters; the skip rate calculation formula is used to indicate whether to skip the current frame, and the skip rate is negatively correlated with the bit rate and positively correlated with the distortion.

[0118] Based on the above embodiments, as a preferred embodiment, it further includes:

[0119] The strategy setting module is used to determine the set selection strategy from the selection strategy pool, which includes at least roulette selection strategy and tournament selection strategy.

[0120] Based on the above embodiments, as a preferred embodiment, it further includes:

[0121] A population operation configuration module is used to determine the population operation type; the population operation type includes heredity, mutation, crossover and replication; and to determine the population operation probability corresponding to each of the population operation types.

[0122] This application also provides an embodiment corresponding to a computer-readable storage medium. The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the method described in the above method embodiments.

[0123] It is understood that if the methods in the above embodiments are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and executes all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0124] The computer-readable storage medium provided in this embodiment includes the method mentioned above, and has the same effect.

[0125] This application also provides an electronic device, see [link to document]. Figure 4 The present application provides a structural diagram of an electronic device, as shown in the embodiment. Figure 4As shown, it may include a processor 1410 and a memory 1420.

[0126] The processor 1410 may include one or more processing cores, such as a quad-core processor or an octa-core processor. The processor 1410 may be implemented using at least one hardware form selected from DSP (Digital Signal Processing), FPGA (Field-Programmable Gate Array), and PLA (Programmable Logic Array). The processor 1410 may also include a main processor and a coprocessor. The main processor, also known as a CPU (Central Processing Unit), is used to process data in the wake-up state; the coprocessor is a low-power processor used to process data in the standby state. In some embodiments, the processor 1410 may integrate a GPU (Graphics Processing Unit), which is responsible for rendering and drawing the content to be displayed on the screen. In some embodiments, the processor 1410 may also include an AI (Artificial Intelligence) processor, which is used to handle computational operations related to machine learning.

[0127] The memory 1420 may include one or more computer-readable storage media, which may be non-transitory. The memory 1420 may also include high-speed random access memory and non-volatile memory, such as one or more disk storage devices or flash memory devices. In this embodiment, the memory 1420 is used to store at least the following computer program 1421, which, after being loaded and executed by the processor 1410, is capable of implementing the relevant steps in the methods executed by the electronic device side as disclosed in any of the foregoing embodiments. In addition, the resources stored in the memory 1420 may also include an operating system 1422 and data 1423, etc., and the storage method may be temporary storage or permanent storage. The operating system 1422 may include Windows, Linux, Android, etc.

[0128] In some embodiments, the electronic device may further include a display screen 1430, an input / output interface 1440, a communication interface 1450, a sensor 1460, a power supply 1470, and a communication bus 1480.

[0129] certainly, Figure 4 The structure of the electronic device shown does not constitute a limitation on the electronic device in the embodiments of this application. In practical applications, the electronic device may include more than [other components]. Figure 4 More or fewer components as shown, or combinations of certain components.

[0130] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. As the system provided in the embodiments corresponds to the method provided in the embodiments, the description is relatively simple; relevant parts can be found in the method section.

[0131] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. It should be noted that those skilled in the art can make several improvements and modifications to this application without departing from the principles of this application, and these improvements and modifications also fall within the protection scope of this application.

[0132] It should also be noted that, in this specification, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

Claims

1. A video encoding method, characterized in that, include: Determine the proxy task; the proxy task is used to characterize the performance of candidate hyperparameter combinations. Define the solution space and fitness function of the genetic algorithm; the solution space of the genetic algorithm is used to define the search range. In the solution space of the genetic algorithm, the fitness function is invoked to map individuals to numerical scores based on the results of the proxy task; the individuals are combinations of hyperparameters. Based on the numerical scores and the set selection strategy, the selected individuals will be placed into the mating pool to form the parent population; The parent population is subjected to population operations in a loop with a set probability until a set number of iterations is met or the proxy task meets a preset threshold. The hyperparameters of the entropy encoding skip mechanism are then output. Video encoding is performed based on the hyperparameters of the entropy coding skipping mechanism.

2. The video encoding method according to claim 1, characterized in that, The task of determining the agent includes: Determine the type of proxy task; the proxy task type includes zero sample, few sample, and many sample. The number of sequence runs corresponding to the hyperparameter combination is determined based on the agent task type.

3. The video encoding method according to claim 1, characterized in that, The process of defining the solution space of the genetic algorithm includes: Determine the key parameters corresponding to chromosomes in the genetic algorithm; the key parameters include the skip function, the basic skip probability constant, and the frame sequence scaling function. Several specific chromosomes are determined based on the key parameters; the solution space of the genetic algorithm consists of all the specific chromosomes.

4. The video encoding method according to claim 3, characterized in that, After defining the solution space and fitness function of the genetic algorithm, the following is also included: The skip rate calculation formula for entropy coding is defined based on the key parameters; the skip rate calculation formula is used to indicate whether to skip the current frame, and the skip rate is negatively correlated with the bit rate and positively correlated with the distortion.

5. The video encoding method according to claim 1, characterized in that, Before the selected individuals are placed into the mating pool to form the parent population based on the numerical scores and the set selection strategy, the following steps are also included: The set selection strategy is determined from a selection strategy pool, which includes at least roulette selection strategies and tournament selection strategies.

6. The video encoding method according to claim 1, characterized in that, Before performing population operations on the parent population cycle with a set probability, the method further includes: Determine the type of population operation; the type of population operation includes heredity, mutation, crossover, and replication; Determine the population operation probability corresponding to each of the aforementioned population operation types.

7. A video encoding system, characterized in that, include: The task determination module is used to determine the proxy task; the proxy task is used to characterize the performance of candidate hyperparameter combinations. The genetic algorithm definition module is used to define the genetic algorithm solution space and fitness function; the genetic algorithm solution space is used to define the search range. The individual scoring module is used to map individuals to numerical scores based on the results of the proxy task by calling the fitness function in the solution space of the genetic algorithm; the individual is a combination of hyperparameters. The parent population generation module is used to select individuals into the mating pool to form the parent population based on the numerical score and the set selection strategy. The iteration module is used to perform population operations on the parent population cyclically with a set probability until a set number of iteration steps is met or the proxy task meets a preset threshold, and output the hyperparameters of the entropy encoding skip mechanism. The video encoding module is used to perform video encoding based on the hyperparameters of the entropy encoding skipping mechanism.

8. An electronic device, characterized in that, include: Memory, used to store computer programs; A processor for executing the computer program to implement the steps of the method as claimed in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed, implements the steps of the method as described in any one of claims 1 to 6.

10. A computer program product, characterized in that, Includes a computer program, which, when executed, implements the steps of the method as described in any one of claims 1 to 6.

Images