Video coding method and system based on multi-channel concurrent software and hardware mixing

Abstract

The invention relates to the technical field of video coding, in particular to a video coding method and system based on multi-channel concurrent software and hardware mixing, and the method comprises the following steps: analyzing brightness distribution and edge structures, screening effective frames, calculating frame priorities, dividing image blocks and combining processing units, counting the number of frame streams, and analyzing frequency concentration. And evaluating image coherence and a load state, and outputting a processing channel adjustment record. According to the invention, through screening of image content features, occupation of redundant frames on coding resources is reduced, through dynamic configuration of frame priorities, timeliness and accuracy of target frame scheduling are improved, an aggregation strategy of a block structure is combined, picture consistency after image processing is enhanced, and task frequency trend identification is utilized, so that image processing efficiency is improved. According to the method, accurate resource matching of a high-load flow section is achieved, joint judgment of image boundary continuity and structure hopping frequency is adopted, an allocation strategy of a processing channel is optimized, an image structure and channel scheduling form linkage, and stability and adaptive capacity are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of video encoding, and in particular to a video encoding method and system based on multi-channel concurrent software and hardware mixing. Background Art

[0002] The field of video coding technology includes various methods and systems for digital compression, transmission and storage of audio and video signals. It aims to convert the original video signal into a compressed format through coding algorithms, reduce data redundancy, and improve transmission and storage efficiency. Video coding involves multiple technical links such as inter-frame and intra-frame prediction, transformation, quantization, entropy coding, etc., including coding control strategies and coding structure design. On this basis, in order to adapt to different application scenarios, video coding technology continues to develop in the direction of high efficiency, low latency, and low power consumption, forming a variety of implementation architectures mainly composed of software codecs, hardware accelerators, and software and hardware collaborative systems.

[0003] Among them, the video encoding method based on multi-channel concurrent software and hardware hybrid refers to the use of multiple channels to concurrently execute encoding tasks during the video encoding process, and to perform encoding processing through software control and collaboration with hardware circuits. It targets the needs of large-scale video data processing and covers multiple aspects of video encoding task division, concurrent execution mechanism, and software and hardware collaborative interface design. Specifically, it divides the original video stream into multiple encoding subtasks according to the granularity of frames or macroblocks, allocates tasks to multiple processing units through a scheduling mechanism, and coordinates the scheduling process and data transmission by the software system, and uses customized hardware circuits to complete computationally intensive subtasks such as transformation, quantization, and entropy coding, forming a software and hardware hybrid concurrent processing structure.

[0004] When faced with concurrent encoding tasks of multiple frame inputs, traditional video coding technology relies on static scheduling executed in timestamp order. It lacks a judgment and response mechanism for image content volatility or target bursts, which may cause important image frames to be processed later. In terms of image structure division, partitioning operations are performed at a fixed granularity, and the structural scale is not dynamically adjusted according to changes in image boundaries, resulting in a poor match between the image partition structure and the actual visual content. At the resource scheduling level, memory allocation is triggered by setting a threshold, and there is a lack of recognition of the dynamic trend of task density changes, resulting in resources fluctuating repeatedly between task concentration overload and sparse segment vacancy. The image channel allocation link adopts fixed or cyclic mode processing, and has no mechanism to identify the continuity or mutation of image boundaries. Delays or encoding errors caused by mismatches in adjacent frame structures may occur. For example, in the monitoring screen, the image coherence of the continuous target moving area is high but still treated equally, resulting in system response distortion and waste of computing resources. Summary of the Invention

[0005] The purpose of the present invention is to solve the shortcomings of the prior art and to propose a video encoding method and system based on multi-channel concurrent software and hardware mixing.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: a video encoding method based on multi-channel concurrent software and hardware hybrid, comprising the following steps:

[0007] S1: Obtain the video frame sequence to be processed, extract the pixel brightness distribution, edge grayscale structure and regional texture contour of each frame image, evaluate the similarity of each pair of adjacent frames, identify redundant frames and filter valid frames, and generate video frame filtering records;

[0008] S2: Based on the video frame screening records, the generation order of valid frames and the dynamic change response characteristics of the corresponding target area are analyzed, the appearance time and the closeness of the content fluctuation of the target frame are identified, and the processing priority of each frame is calculated by comparing the distribution density of multiple image frames on the time axis to generate a task level mapping mark;

[0009] S3: calling the task level mapping marker, dividing each frame of the image into multiple blocks, analyzing the degree of pixel brightness change in the same block between adjacent frames, evaluating the continuity of the edge structure between blocks, combining the image blocks into processing units, and generating an image structure aggregation result;

[0010] S4: Call the image structure aggregation result, obtain the continuous input of the video frame sequence, extract the number of frame flows in a continuous time period, analyze the concentration of the arrival frequency of encoding tasks, evaluate the load status, screen the task flows that need to allocate exclusive buffer areas, and generate the encoding task buffer level.

[0011] As a further solution of the present invention, the video frame screening record includes frame continuity judgment results, redundant frame index list, and valid frame sequence; the task level mapping mark includes frame processing level number, priority assignment label, and event trigger sorting result; the image structure aggregation result includes block merging mapping relationship, structural connectivity partition, and intra-block edge structure indication; the encoding task buffer level includes task flow scheduling label, buffer usage ratio information, and frame flow density distribution information.

[0012] As a further solution of the present invention, the steps of obtaining the video frame screening record are specifically as follows:

[0013] S111: Obtain a sequence of video frames to be processed, call pixel distribution data of each frame image, analyze the brightness distribution state, edge grayscale structure trend and texture contour coding mode of each area in the image, and generate basic feature data of the frame image;

[0014] S112: Based on the basic feature data of the frame images, analyzing the brightness variation range and grayscale structure connection characteristics of adjacent frames in the same area, evaluating the structural stability and texture disturbance intensity between consecutive images, calculating the regional similarity measurement value between adjacent frames, and obtaining the adjacent frame similarity measurement value;

[0015] S113: calling the adjacent frame similarity metric value, marking redundant frames and adding frame skipping identifiers according to the similarity, screening valid frames and constructing a valid frame sequence, and generating a video frame screening record.

[0016] As a further solution of the present invention, the step of obtaining the task level mapping mark is specifically as follows:

[0017] S211: Based on the video frame screening record, analyzing the generation order of valid frames in the continuous sequence, obtaining the frame index number and the position relationship between each image frame, and generating a valid frame generation time sequence feature;

[0018] S212: Calling the valid frame to generate a temporal feature quantity, analyzing the continuous response characteristics of the target area in each frame image, identifying the position change and stay of the target area in the continuous image, extracting the continuous number of the area in the motion path and analyzing the concentration state of the change, and generating a target area response density index;

[0019] S213: Analyze the central tendency of the image frames in content change and temporal distribution according to the target area response density index, calculate the processing priority score of each frame, construct a processing task sequence, and establish a task level mapping mark.

[0020] As a further solution of the present invention, the steps of obtaining the image structure aggregation result are specifically as follows:

[0021] S311: Based on the task level mapping mark, the current frame is divided into multiple image blocks, and pixel brightness data of corresponding blocks in the current frame and the previous frame are analyzed to identify the degree of pixel brightness change, evaluate the magnitude of the change in image content, and generate a brightness change difference data set;

[0022] S312: Calling the brightness change difference dataset, analyzing the edge pixel color direction of each pair of adjacent image blocks in the current frame, identifying the consistency of the direction angle, analyzing the continuity of the edge color, and generating edge direction consistency information;

[0023] S313: Pair and combine image blocks according to the edge direction consistency information, compare the brightness change trend and edge direction consistency between blocks, calculate the merging adaptation coefficient, combine multiple image blocks into independent processing units, and establish an image structure aggregation result.

[0024] As a further solution of the present invention, the step of obtaining the encoding task buffer level is specifically as follows:

[0025] S411: Calling the image structure aggregation result, obtaining continuous input data of the video frame sequence, analyzing the number of video frames received in each time period, and generating a frame flow statistics record for each time period;

[0026] S412: Calculating the difference between the number of frame flows in each pair of adjacent time periods based on the frame flow statistics for the time period, identifying the continuity of the direction of change in the number of frame flows, analyzing the concentration of the arrival frequency of the encoding task, and generating a frequency concentration distribution result;

[0027] S413: Call the frequency concentration distribution result, filter the task flow that needs to be allocated an exclusive buffer area, and generate the encoding task buffer level.

[0028] As a further embodiment of the present invention, the method further comprises:

[0029] S5: Calling the encoding task buffer level and image structure aggregation results, extracting boundary grayscale change information of multiple partitions in the current frame, evaluating the consistency and coherence of the overall image transition by analyzing the edge continuity and grayscale stability between partitions, comparing the connection strength of internal partitions of the image and the external jump distribution frequency, adjusting the processing channel of each image frame, and generating a processing channel adjustment record;

[0030] The processing channel adjustment record includes channel allocation instructions, inter-frame channel change sequence, and channel load association mapping information.

[0031] As a further solution of the present invention, the step of obtaining the processing channel adjustment record is specifically as follows:

[0032] S511: Calling the encoding task buffer level and image structure aggregation result, extracting the boundary pixel grayscale information of each image partition in the current frame, recording the grayscale change direction and amplitude between adjacent partitions, analyzing the grayscale continuous change characteristics at the boundary position, and generating a boundary grayscale change data set;

[0033] S512: Analyze the consistency of edge grayscale continuity between image partitions based on the boundary grayscale change dataset, evaluate the stability of grayscale change amplitudes in adjacent regions, identify the transition connection strength of each frame in spatial distribution, and generate an image coherence score;

[0034] S513: Calling the image coherence score, comparing the connection strength between each partition in the current frame with the grayscale jump distribution of the frame edge area, adjusting the processing channel of each image frame, and establishing a processing channel adjustment record.

[0035] A video encoding system based on multi-channel concurrent software and hardware hybrid is used to execute the above-mentioned video encoding method based on multi-channel concurrent software and hardware hybrid, and the system includes:

[0036] The redundant frame filtering module extracts pixel brightness features, performs edge grayscale structure analysis, and detects regional texture distribution for each frame based on the video frame sequence to be processed. It also analyzes the similarity of each pair of adjacent frames, identifies redundant frames, extracts valid frames, and generates video frame screening records.

[0037] The task sorting module screens the video frames and analyzes the generation order of valid frames and the dynamic change response characteristics of the corresponding target area based on the video frame screening record, identifies the appearance time of the target frame and the closeness of the content fluctuation, calculates the processing priority of each frame by comparing the distribution density of multiple image frames on the time axis, and establishes a task level mapping mark;

[0038] The region combination module divides the image frame into blocks based on the task level mapping mark, analyzes the degree of pixel brightness change in the same block between adjacent frames, evaluates the continuity of the edge structure between blocks, combines the blocks and constructs independent processing units to establish the image structure aggregation result;

[0039] The buffer allocation module collects the video frame flow in a continuous period based on the image structure aggregation result, analyzes the concentration of the encoding task arrival frequency, evaluates the load status in real time, identifies the task flow that needs to be allocated an exclusive buffer area, and establishes the encoding task buffer level;

[0040] Based on the encoding task buffer level and image structure aggregation results, the channel optimization module detects and records the direction of grayscale change at the partition boundary within each image frame, analyzes the stability of the grayscale transition at the edges between adjacent partitions, evaluates the partition connection strength of the entire frame image, adjusts the processing channel of each image frame, and establishes a processing channel adjustment record.

[0041] Compared with the prior art, the advantages and positive effects of the present invention are:

[0042] In the present invention, the occupation of coding resources by redundant frames is reduced by screening image content features, the timeliness and accuracy of target frame scheduling are improved by dynamically configuring frame priorities, the aggregation strategy of block structure is combined to enhance the consistency of the image after image processing, the task frequency trend is identified to achieve accurate resource matching of high-load flow segments, the joint judgment of image boundary continuity and structural jump frequency is adopted to optimize the allocation strategy of processing channels, so that the image structure and channel scheduling are linked, and the stability and adaptability in complex concurrent scenarios are improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1It is a schematic diagram of the workflow of the present invention;

[0044] Figure 2 A flowchart for obtaining video frame screening records of the present invention;

[0045] Figure 3 Obtaining a flow chart for task level mapping tags of the present invention;

[0046] Figure 4 A flowchart for obtaining image structure aggregation results of the present invention;

[0047] Figure 5 A flowchart for obtaining the buffer level of an encoding task according to the present invention;

[0048] Figure 6 A flow chart for obtaining a record of adjusting the processing channel of the present invention. DETAILED DESCRIPTION

[0049] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0050] In the description of the present invention, it should be understood that the terms "length," "width," "up," "down," "front," "back," "left," "right," "vertical," "horizontal," "top," "bottom," "inside," "outside," and the like, indicating positions or relationships, are based on the positions or relationships shown in the accompanying drawings and are intended only to facilitate the description of the present invention and simplify the description. They do not indicate or imply that the devices or elements referred to must have a specific orientation, be constructed, or operate in a specific orientation. Therefore, they should not be construed as limiting the present invention. Furthermore, in the description of the present invention, "plurality" means two or more, unless otherwise expressly and specifically defined.

[0051] See also Figure 1 The present invention provides a technical solution: a video encoding method based on multi-channel concurrent software and hardware hybrid, comprising the following steps:

[0052] S1: Obtain the video frame sequence to be processed, extract the pixel brightness distribution, edge grayscale structure and regional texture contour of each frame image, evaluate the similarity of each pair of adjacent frames, identify redundant frames and filter valid frames, and generate video frame filtering records;

[0053] S2: Based on the video frame screening records, the generation order of valid frames and the dynamic change response characteristics of the corresponding target area are analyzed. The appearance time and content fluctuation density of the target frame are identified. By comparing the distribution density of multiple image frames on the time axis, the processing priority of each frame is calculated and a task level mapping mark is generated.

[0054] S3: Call the task-level mapping marker to divide each frame into multiple blocks, analyze the degree of pixel brightness change in the same block between adjacent frames, evaluate the continuity of the edge structure between blocks, combine the image blocks into processing units, and generate the image structure aggregation result;

[0055] S4: Call the image structure aggregation results to obtain the continuous input of the video frame sequence, extract the number of frame flows in a continuous time period, analyze the concentration of the encoding task arrival frequency, evaluate the load status, screen the task flow that needs to be allocated an exclusive buffer area, and generate the encoding task buffer level;

[0056] S5: Call the encoding task buffer level and image structure aggregation results, extract the boundary grayscale change information of multiple partitions in the current frame, evaluate the consistency and coherence of the overall image transition by analyzing the edge continuity and grayscale stability between partitions, compare the connection strength of the internal partitions of the image and the external jump distribution frequency, adjust the processing channel of each image frame, and generate a processing channel adjustment record.

[0057] Video frame screening records include frame continuity judgment results, redundant frame index list, and valid frame sequence. Task level mapping marks include frame processing level number, priority allocation label, and event trigger sorting results. Image structure aggregation results include block merging mapping relationship, structural connectivity partition, and intra-block edge structure indication. Encoding task buffer level includes task flow scheduling label, buffer usage ratio information, and frame flow density distribution information. Processing channel adjustment records include channel allocation instructions, inter-frame channel change sequence, and channel load association mapping information.

[0058] See also Figure 2 ,The specific steps for obtaining video frame screening records are:

[0059] S111: Obtain a sequence of video frames to be processed, call pixel distribution data of each frame image, analyze the brightness distribution state, edge grayscale structure trend and texture contour coding mode of each area in the image, and generate basic feature data of the frame image;

[0060] First, an input sequence of several consecutive frames is obtained from the external video capture module, and the time index number is marked respectively. For example, a certain video frame sequence is obtained as 10 frames numbered from t=1 to t=10. Then, the pixel brightness data of each frame is extracted. Taking the first frame as an example, the image is divided into regions of uniform size, such as 4 horizontal blocks and 4 vertical blocks, for a total of 16 partitions. For the partition number i=5, the brightness of all pixels in the partition is counted. Assuming that the number of pixels in the region is 256 and the brightness range of each pixel is [0, 255], the pixel brightness values are 200, 198, 205, etc., and the average value is normalized to [0, 1]. For example, if the average value is 198.5, the brightness normalization value is Similarly, the normalized brightness mean of other areas is calculated to form 16 partition brightness data sets and recorded in the data structure. The edge grayscale structure direction is calculated by calculating the gradient direction of the partition edge pixel gradient vector. Taking the gradient direction angles of the pixel points at the partition boundary (such as points numbered 1 to 16 on the boundary line) as an example, they are 25°, 28°, etc., respectively, and recorded as the partition edge grayscale structure direction data. The texture contour is extracted by using the texture direction. The number of features with a texture direction of 30° is 80 points, the number of features with a direction of 60° is 60 points, and the number of features with a direction of 90° is 116 points. The maximum value is mapped to the interval [0, 1] range, that is, The texture data of each region is counted to obtain the texture coding mode of 16 regions in each frame image. The above steps are repeated until the 10th frame is processed to obtain 10 sets of frame image basic feature data sets. Finally, the processed data are classified and stored as frame image basic feature data.

[0061] S112: Based on the basic feature data of the frame image, analyze the brightness variation range and grayscale structure connection characteristics of adjacent frames in the same area, and evaluate the structural stability and texture disturbance intensity between consecutive images using the formula:

[0062]

[0063] Calculate the region similarity measure between adjacent frames and obtain the adjacent frame similarity measure;

[0064] Among them, S f (t, t-1) is the region similarity measure between the current frame and the previous frame, which represents the joint scoring result of the structure and brightness change between the two frames. is the normalized mean value of the pixel brightness of the i-th image partition in the current frame, which is obtained by counting the grayscale values of all pixels in the partition and dividing it by the upper limit of the maximum grayscale range. is the normalized mean value of the pixel brightness of the i-th image partition in the previous frame, is the grayscale continuity index of the current frame and the previous frame at the boundary of the i-th image partition, which is obtained by extracting the gradient direction of the edge pixels between adjacent frames and calculating the cosine similarity of the angle. is the normalized encoding value of the direction structure of the j-th texture region in the current frame, obtained by the texture direction extraction algorithm and then mapping the angle to the interval [0,1].

[0065] is the directional structure normalized encoding value of the j-th texture region in the previous frame, n represents the total number of regions into which the image is divided, m represents the number of texture analysis blocks delineated, ε is the image structure perturbation adjustment term, which is set to the normalized root mean square value of the texture perturbation residual, obtained by normalizing the root mean square of the texture direction difference in multiple frame pairs, to avoid unstable calculation caused by a denominator of zero or too small, t is the temporal index number of the current frame, t-1 is the index number of the previous frame in the sequence, i is the image partition number index in the brightness and edge continuity calculation, and j is the texture region number index in the texture structure direction normalization calculation;

[0066] According to the obtained data sequence, two adjacent frames such as the first frame (t-1=1) and the second frame (t=2) are taken for analysis. Taking the area numbered 5 as an example, the normalized brightness value of the first frame is called and the normalized brightness value of the second frame Calculate the absolute value of the brightness difference The continuity calculation of the edge grayscale structure is to extract the gradient vector angle at the edge of region 5. Assuming that the average angle of the edge gradient direction of the first frame is 27°, the average angle of the corresponding area of the second frame is 29°, and the angle is 2°, the cosine similarity of the angle is calculated to be cos(2°) = 0.999, which is recorded as the edge continuity index. Then, the brightness difference of the region is multiplied by the edge continuity index to obtain the brightness edge combination index of a single region, which is 0.007×0.999=0.006993. All 16 regions are calculated one by one and summed. Taking the total number of regions n=16 as an example, the calculated sum is 0.078. According to the texture coding data, the texture direction normalized coding values of region number 3 are respectively the first frame Second frame Subtracting the square yields (0.570-0.550) 2 =0.0004. Similarly, the calculation for other texture blocks j (m=16) is completed and the sum is assumed to be 0.0056. The square root is then taken as 0.0748. The disturbance adjustment term ε is the normalized result of the standard deviation of the average texture difference obtained by statistics of multiple adjacent frames in the training set of different video samples. The experimental value is 0.005 to prevent the denominator of the calculation process from being zero or close to zero. The calculation formula is:

[0067]

[0068] Substitute the set values for calculation:

[0069]

[0070] Among them, the regional similarity measurement value represents the overall content similarity score between two consecutive frames at the image region level, reflecting the joint similarity characteristics of the current frame and the previous frame in terms of local brightness changes, structural boundary continuity, and texture stability. The larger the value, the smaller the change in the image region and the more stable the structure, which can be regarded as a redundant frame. The smaller the value, the presence of global disturbances or structural jumps, which may contain key event information. This parameter is used to drive the trigger mechanism of frame skipping marking to improve the encoding efficiency and resource allocation accuracy of the video processing system in multi-channel concurrent scenarios. In continuous video frames, there are a large number of static backgrounds or slowly changing areas. In order to save encoding resources, it is necessary to identify content-redundant frames, that is, to determine whether the current frame is highly similar to the previous frame. The formula implements this judgment through three dimensions, namely, brightness changes are used to measure content differences at the pixel level, boundary grayscale continuity identifies whether the image structure is coherent, and texture direction perturbation detects whether there is a structural jump in the global picture, ensuring that the formula identifies local significant changes and suppresses the risk of misjudging a valid frame due to overall texture oscillation. The score value provides an intuitive and comparable input basis for the frame skipping control module, realizing encoding task frame screening and channel resource scheduling driven by real image content features. In the formula, the numerator is the sum of the absolute value of the brightness change of each partition multiplied by the edge continuity of the corresponding area to reflect the degree of union of adjacent frames in terms of brightness and structural changes; the denominator is used to measure the perturbation intensity of the texture structure. The square root of the sum of the squares of the texture differences is added and the adjustment term is added to avoid instability in division. The final calculation result S f The value interval of (2,1)=0.978 is [0, 1], where the interval [0.9, 1] is determined to be highly similar, [0.6, 0.9] is generally similar, and [0, 0.6] is dissimilar or has significant changes. The value 0.978 indicates that the first and second frames are highly similar.

[0071] S113: Calling the similarity metric of adjacent frames, marking redundant frames and adding frame skipping flags based on the similarity, screening valid frames and constructing a valid frame sequence, and generating a video frame screening record;

[0072] Call the similarity metric of adjacent frames. For example, the similarity between the second frame and the first frame is 0.978, which exceeds the set similarity threshold T. s=0.95. The threshold value is set based on the statistical distribution of similarity between adjacent frames of a large number of video samples. The sample size is 1000 sets of video frames. The average similarity value is calculated to be 0.82 and the standard deviation is 0.08. Based on the actual visual effect and data statistics of normal video playback, the threshold value is selected as 0.95, which is higher than the average value and one and a half standard deviations to ensure that the redundant frames screened out have extremely high similarity. Therefore, the second frame is determined to be a redundant frame and marked as a skipped frame. If the similarity between the third frame and the second frame is calculated, S f (3,2)=0.89 does not reach the threshold, so the third frame is a valid frame and is added to the valid frame sequence. This method is used to process subsequent frames one by one. Assuming that 10 frames have been processed, frame numbers 1, 3, 5, 7, and 9 are judged to be valid, while frames 2, 4, 6, 8, and 10 are redundant. Finally, a valid frame sequence {1, 3, 5, 7, 9} is established, and corresponding screening records are generated, as shown in Table 1.

[0073] Table 1 Video frame screening record table

[0074] Frame Number Similarity Is it effective? 1 / yes 2 0.978 no 3 0.89 yes 4 0.96 no 5 0.87 yes 6 0.97 no 7 0.85 yes 8 0.98 no 9 0.88 yes 10 0.97 no

[0075] As shown in Table 1, the similarity calculated for each frame is compared with the threshold T s =0.95, judge whether each frame is valid and mark the filter record to establish a valid frame sequence.

[0076] See also Figure 3 , the specific steps for obtaining the task level mapping mark are:

[0077] S211: Based on the video frame screening record, analyze the generation order of valid frames in the continuous sequence, obtain the frame index number and the position relationship between each image frame, and generate the valid frame generation time sequence feature;

[0078] Using the video frame screening record, based on the valid frame sequence screened out by the previous stage, it is assumed that the index numbers of the valid frames in a set of video input frame sequences are 1, 3, 5, 7, and 9 respectively. The index numbers of these frames are arranged and stored in the order of video input. For example, the index number of the first frame is 1, the index number of the second frame is 3, and so on to form a sequence record. Then, by traversing each valid frame position, the front and back position relationship of the valid frames in the original video sequence is judged one by one. Taking the frame with frame index number 5 as an example, by comparing its adjacent valid frame index numbers 3 and 7, the number of position intervals is calculated respectively. The frame interval number between frame 5 and frame 3 is 1 frame, and the interval number between frame 5 and frame 7 is also 1 frame. The position relationship is recorded in the effective frame generation time series feature data. For example, the time series feature of frame 5 is that the spacing between the front and back frames is 1. Similarly, the spacing data of other valid frames are calculated to construct a time series feature table, as shown in Table 2.

[0079] Table 2 Effective frame generation timing feature table

[0080]

[0081]

[0082] As shown in Table 2, the effective frame generation time sequence feature is generated by calculating the relationship between the index number of the effective frame and the spacing between the previous and next adjacent effective frames.

[0083] S212: Calling valid frames to generate temporal feature quantities, analyzing the continuous response characteristics of the target area in each frame image, identifying the position change and stay of the target area in the continuous images, extracting the continuous number of the area in the motion path and analyzing the concentration of the change, and generating the target area response density index;

[0084] Taking the 5th frame as an example, the response characteristics of the target area in each frame are analyzed one by one. Assuming that a target area in the 5th frame is defined as the position of area number 6, its continuous response state is detected in consecutive image frames. For example, the position coordinates of the corresponding area 6 in the 3rd, 5th and 7th frames are (150, 200), (152, 202) and (155, 205) respectively. Then the position change of the target area in the consecutive frames is the Euclidean distance of the position coordinate difference between the 3rd and 5th frames. Euclidean distance of the coordinate difference between the 5th frame and the 7th frame The cumulative calculated target area position change is 2.83+4.24=7.07, and then the target area's stay in consecutive frames is counted. If the position coordinate difference is set to be less than the threshold δ=1.5 pixels, it is judged to be staying, and the position coordinate difference is greater than δ, it is judged to be changing. Region 6 is continuously in a changing state between the 3rd and 7th frames and does not stay. According to the position change and stay state, the region continuous response feature data is recorded. For example, the continuous response sequence of region number 6 is marked as a continuous change state. Other target regions are processed similarly. Taking a total of 16 target regions in frame 5 as an example, the change concentration state of the continuous response state of all regions is counted. Assuming that a total of 12 regions change continuously, the continuous change ratio is 12 / 16=0.75, that is, the response density index is 0.75. In this way, the effective frame response density index is analyzed frame by frame to generate the target region response density index data table, as shown in Table 3.

[0085] Table 3 Target area response density index table

[0086] Frame index number Number of continuous response areas Response Intensity Index 1 10 0.625 3 11 0.687 5 12 0.750 7 13 0.812 9 12 0.750

[0087] As shown in Table 3, the target area response density index is calculated by counting the number of continuous response areas.

[0088] S213: Analyze the central tendency of the image frame in terms of content change and temporal distribution based on the target area response density index, using the formula:

[0089]

[0090] Calculate the processing priority score of each frame image, build a processing task sequence, and establish a task level mapping mark;

[0091] Among them, P l is the processing priority score of the lth frame, which is used to determine the task sorting number in the encoding schedule, A′ l is the normalized value of the response time of the target area in the first frame, which indicates the length of time that the target area keeps moving or changing in the continuous frames. It is obtained by identifying the continuous response sequence of the same position area in adjacent frames and dividing it by the maximum response time. l,k is the normalized value of the grayscale jump intensity of the kth image partition in the lth frame, which represents the change amplitude of the grayscale value of the partition in this frame and the grayscale value of the same position in the previous frame. It is calculated by dividing the brightness difference of the pixels in the same partition between adjacent frames by the maximum grayscale difference. l is the normalized index of the frequency domain change effect of the lth frame, which represents the content energy concentration of the edge area in the frequency dimension of the frame. It is obtained by performing fast Fourier transform on the edge area to obtain the spectrum intensity and then normalizing it. S l is the number of structurally connected regions in the lth frame, indicating the number of interconnected regional blocks in the image partition, which is obtained by marking and counting through the edge detection and regional connectivity analysis process. l The index number of the lth frame in the valid frame sequence indicates the position of the frame in the current encoding cycle and can be directly obtained through the sequential number. It is the average value of all valid frame index numbers in the current coding cycle, indicating the central reference position of the task time period. l Find the arithmetic mean and get W l is the texture jump ratio of the lth frame image, which indicates the area ratio of the region with sudden texture changes in the frame. It is obtained by counting the number of pixels in the region where the gradient direction changes sharply in the image and dividing it by the total number of pixels in the image. l is the number index of the image frame, which indicates the order of the current processing frame in the valid sequence. k is the number index of the image partition, which indicates the position of the sub-region after the image is divided. M is the total number of partitions divided into each frame image, which is a preset fixed parameter.

[0092] Taking the 5th frame as an example for analysis, according to the response density index 0.750 obtained above, the processing task order is determined by calculating the processing priority score. First, the normalized value A′5 of the response time of the 5th frame is extracted. For example, the target area of the 5th frame changes continuously from the 3rd frame to the end of the 7th frame, and experiences 2 intervals (3 to 5 frames, 5 to 7 frames). Assuming that the maximum response duration interval is 5 frames, then A′5=2 / 5=0.4. Taking the grayscale jump intensity as an example, the grayscale difference between the first area of the 5th frame and the grayscale value change of the corresponding area of the 3rd frame is analyzed. For example, the brightness of the first area in the 3rd frame is 0.60, and the brightness of the 5th frame is 0.75. Then the grayscale jump intensity value is |0.75-0.60|=0.15. The normalization is based on the maximum grayscale difference of 0.5, then B′ 5,1 =0.15 / 0.5=0.3. Similarly, the sum of the jump strengths of all 16 regions is analyzed and assumed to be 2.4. The frequency domain change normalization index Z′5 is 0.35 based on the concentrated statistics of the spectrum intensity after FFT in the edge area. The number of structural connected regions S5 is assumed to be 8. The index number of the 5th frame in the valid frame sequence is D5=5. The average value of the index numbers of all valid frames is The absolute value of the index difference is The texture jump ratio W5 is calculated as the area of the texture direction mutation region in the image accounting for 0.2 of the total image area. It is used to calculate the processing priority score:

[0093]

[0094] Substituting the values:

[0095]

[0096] The processing priority score refers to the task scheduling ranking weight for each frame in the current encoding cycle. It is a numerical indicator calculated based on multiple dimensions such as the image frame's temporal change trend, structural fluctuation characteristics, frequency domain activity, inter-frame temporal distribution density, and texture jump degree. The larger the score, the more urgent the frame's response to the encoder task, and the earlier it should be sent to the encoding execution process. The parameters are used to guide frame-level task sorting, encoding resource allocation, and concurrent path selection. This ensures that key image frames are responded to in a timely manner, avoids resources being occupied by a large number of redundant or low-timeliness frames, and achieves content-aware optimization of task execution in high-concurrency scenarios. The formula comprehensively reflects the importance of the frame by integrating the state of the frame image in three different dimensions: the temporal response state (whether the target area changes continuously for a long time), the spatial structure state (whether the local area has strong structural changes), and the frequency domain energy state (whether the frequency characteristics of the image edge are concentrated). The structural complexity and temporal deviation terms in the denominator are used to identify frames that, despite strong changes, exhibit chaotic structure or appear in non-critical locations, reducing their scheduling priority. Multiplying by the texture jump rate allows for a moderate improvement in frames that are generally stable but exhibit sudden texture fluctuations. This implements a hierarchical processing principle that prioritizes dynamic change capture, weights structural coherence regulation, and enhances texture jumps. This is a fusion response mechanism for the evolution of image content and the rational allocation of scheduling resources. The calculated processing priority score is 0.223, and the priority range is defined as: [0, 0.3] low priority, [0.3, 0.7] medium priority, [0.7, 1] high priority, with 0.223 being in the low priority range. A similar method is used to calculate the processing priority scores for all valid frames. A task level mapping marker data sequence is established based on the numerical values, completing the construction of the processing task sequence.

[0097] See also Figure 4 , the steps to obtain the image structure aggregation results are as follows:

[0098] S311: Based on the task level mapping mark, the current frame is divided into multiple image blocks, and the pixel brightness data of the corresponding blocks in the current frame and the previous frame are analyzed to identify the degree of pixel brightness change, evaluate the magnitude of the change in image content, and generate a brightness change difference dataset;

[0099] Call the task level mapping tag, take the current frame (numbered as frame 5) as the object, first divide the image into multiple even blocks, for example, 4 horizontally and 4 vertically, a total of 16 blocks, then call the pixel brightness data of the corresponding blocks in the 5th frame and the previous frame (numbered as frame 3) one by one. For example, take block 8 as an example, the average pixel brightness of this block in the 5th frame is 180, and the average brightness of this block in the previous frame is 160. The absolute difference is calculated as |180-160|=20. Then, normalization is performed using the maximum absolute value of the brightness change of all block pixels in the current frame (for example, 50), and the normalized brightness change difference value of block 8 is obtained as 20 / 50=0.4, which is stored as the brightness change difference data of block 8. Similarly, the calculation is completed block by block to obtain a total of 16 block brightness change difference data sets. For example, the block brightness change difference data are 0.4, 0.2, 0.6, etc., respectively, forming a complete data record, which is finally recorded as the brightness change difference data set.

[0100] S312: Calling the brightness change difference dataset, analyzing the edge pixel color direction of each pair of adjacent image blocks in the current frame, identifying the consistency of the direction angle, analyzing the continuity of the edge color, and generating edge direction consistency information;

[0101] According to the brightness change difference data set, taking the edge pixel analysis of adjacent image blocks in the current frame as an example, the gradient direction analysis is performed on the edge pixels of each pair of adjacent blocks. Taking block numbers 4 and 5 as an example, the gradient direction of the edge pixel points on the boundary line of the two blocks is extracted. For example, there are a total of 5 edge pixel points on the boundary line, and the gradient direction angles are 30°, 33°, 32°, 35°, and 31° respectively. The consistency analysis of these angle data is performed, and the difference between the direction of each edge pixel and the average angle of the block edge direction is calculated. First, the average value is calculated as Then calculate the difference between the direction angle of each pixel and the average value one by one. For example, the angle difference of the first pixel is |30-32.2|=2.2°. Similarly, calculate all differences and count the number of differences less than the threshold (such as the threshold is 5°). Assuming that all 5 differences are less than 5°, the direction consistency ratio is 5 / 5=1.0, indicating complete consistency. Similarly, complete the edge direction consistency analysis between other adjacent blocks one by one, and finally record the edge direction consistency data between all adjacent blocks to form edge direction consistency information.

[0102] S313: Pair and combine the image blocks based on the edge direction consistency information, and compare the brightness change trend and edge direction consistency between the blocks using the formula:

[0103]

[0104] Calculate the merge adaptation coefficient, combine multiple image blocks into independent processing units, and establish the image structure aggregation result;

[0105] Among them, P i′ is the merged adaptation coefficient of the i′th image block, is the normalized value of the pixel brightness difference between the current frame and the previous frame of the i′th image block, which is obtained by dividing the absolute value of the brightness difference by the maximum brightness change range in the current frame. i′q is the color direction angle of the qth edge pixel in the i′th image block, which is obtained by extracting the angle value of the edge pixel in the image gradient direction. is the average value of the color direction angles of all edge pixels in the i′th image block. i′q The arithmetic mean is C i′ The continuous score of the edge structure of the i′th image block is obtained by calculating the consistency ratio of the gradient direction angles of adjacent edge pixels, R i′ is the channel mapping coefficient of the region where the i′th image block is located, which is obtained by setting the image region structure coherence scoring rule in the system. i′ is the number index of the image block in the current processing frame, which is used to identify each independent image region. q is the serial number index of the edge pixel in the current image block, which is used to traverse all edge pixels in the block. Q is the total number of edge pixels in each image block that participate in the edge direction calculation.

[0106] Extract edge direction consistency information. Taking block number 8 as an example, perform block pairing combination analysis based on its brightness change trend and edge direction consistency. For example, the normalized value of the brightness change difference of block 8 is There are a total of Q=5 edge pixels in this block, with direction angles of 20°, 22°, 19°, 21°, and 23° respectively. The average direction angle is calculated as The absolute values of the differences between the angle of each pixel and the average value are 1°, 1°, 2°, 0°, and 2°, respectively, and the sum is 6°. The sum of the absolute differences in edge direction consistency is 6°. At the same time, the block edge structure continuity score C8 is calculated by the statistical calculation of the gradient angle difference of adjacent edge pixels. Assuming that the obtained angle consistency ratio is 0.85, then C8 = 0.85. The regional channel mapping coefficient R8 is a parameter set according to the system's internal structural coherence scoring rules, with a value range of [0.5, 1.0]. Assuming that according to the scoring rules, the regional structural coherence of block 8 is high, the mapping coefficient value is R8 = 0.9. Substitute into the formula:

[0107]

[0108] Substitute the specific values in block 8 for calculation:

[0109]

[0110] Among them, the merge adaptation coefficient is a multi-factor joint evaluation result of the image block in terms of temporal brightness stability, edge direction continuity and structural connectivity. The larger the value, the more the image block meets the conditions for merging with adjacent blocks into a processing unit. It is used to provide an objective and quantitative decision-making basis for determining which image blocks can be aggregated in image coding tasks, effectively reduce the repeated calculation of redundant boundaries, reduce the computational burden in the coding process, and improve the system's coding efficiency and concurrent processing capabilities in frames with slow image changes or strong regional coherence, ensuring that the merging operation occurs in areas with continuous image structure and stable content, avoiding compression distortion or efficiency reduction caused by coding fragments crossing structural boundaries. The formula comprehensively measures the coordinated relationship between the temporal stability (measured by brightness variation), spatial consistency (measured by color directional distribution), and structural coherence (measured by edge continuity) of a block. If an image region has no significant brightness fluctuations between consecutive frames, its edge structure direction is basically consistent, and there is no sudden change in adjacent pixels, it is judged to be suitable for merging with adjacent images to form a processing unit. The repulsion of unstable areas is enhanced by accumulating the variation terms, while the merging behavior is suppressed at the fracture boundary by the structural factor. Ultimately, the merging behavior is controlled to change dynamically according to the dynamic degree and structural complexity of the actual image. The calculated merging adaptation coefficient of block 8 is 4.388, and the value range is defined as: [0, 2] low adaptation, [2, 5] medium adaptation, and [5, 8] high adaptation. 4.388 is at the medium adaptation level, so block 8 is paired with adjacent blocks with higher edge direction consistency. Assuming that the edge direction consistency of block 8 and block 7 is 0.95, block 8 and block 7 are paired to form an independent processing unit. Similarly, the merging adaptation coefficients of all blocks are calculated one by one, and the image structure aggregation result data record is established, as shown in Table 4.

[0111] Table 4 Image structure aggregation results

[0112] Block number Combined fitness coefficient Adaptability level Pairing Block 1 3.120 Medium Adaptation 2 2 1.800 Low adaptation none 3 5.900 High adaptability 4 4 5.750 High adaptability 3 5 2.400 Medium Adaptation 6 6 2.800 Medium Adaptation 5 7 4.200 Medium Adaptation 8 8 4.388 Medium Adaptation 7

[0113] As shown in Table 4, the blocks are matched according to the calculated merging adaptation coefficients to determine the corresponding pairing relationships and construct the image structure aggregation results.

[0114] See also Figure 5 ,The steps for obtaining the encoding task buffer level are as follows:

[0115] S411: Calling the image structure aggregation result, obtaining continuous input data of the video frame sequence, analyzing the number of video frames received in each time period, and generating a frame flow statistics record for the time period;

[0116] Using the image structure aggregation results, the video frame sequence is taken as input to obtain video frame data for multiple consecutive time periods. Taking the time period length as 1 second as an example, the video frame sequence recorded at the system input end is called. For example, the number of frames received by the system in 5 consecutive time periods are 150 frames, 155 frames, 148 frames, 160 frames and 152 frames respectively. The input video frames are counted and recorded in each time period to form a time period frame flow statistics record. The specific calculation method is to directly count the video frames in each time period. For example, the number of frames counted in the 1st second time period is 150 frames, the number of frames counted in the 2nd second time period is 155 frames, and so on. Finally, a complete time period frame flow statistics record is obtained, as shown in Table 5.

[0117] Table 5 Time period frame flow statistics record table

[0118] Time period number Time period start and end time (s) Frame number (frame) 1 0～1 150 2 1～2 155 3 2～3 148 4 3～4 160 5 4～5 152

[0119] As shown in Table 5, by counting the number of video frames input in each time period, a time period frame flow statistics record is formed.

[0120] S412: Based on the frame flow statistics of each time period, the difference between the number of frame flows in each pair of adjacent time periods is calculated to identify the continuity of the change direction of the frame flow number, analyze the concentration of the encoding task arrival frequency, and generate a frequency concentration distribution result;

[0121] Taking the continuity analysis of the change in the number of frame flows in adjacent time periods as an example, the frame flow number data of adjacent time periods are called to calculate the frame flow difference. Taking the first and second time periods as examples, the frame numbers are 150 frames and 155 frames respectively, then the frame flow difference is 155-150=+5 frames, which represents an increasing trend in the number of frame flows; then analyzing the second to third time periods, the frame numbers are 155 frames and 148 frames respectively, and the calculated difference is 148-155=-7 frames, which represents a decreasing trend in the number of frame flows; the frame flow differences of subsequent time periods are calculated in sequence and the continuity of the change direction is recorded. For example, the change from the third to the fourth time period is 160-148=+12 frames, which continues to show an increasing trend, and the change from the fourth to the fifth time period is 152-160=-8 frames, which turns to a decreasing trend; the continuous The increase or decrease trend is recorded. For example, in this example, the frame flow change trend is recorded as: "increase → decrease → increase → decrease". Through this trend analysis, the order of magnitude of the frame flow changes in adjacent time periods is statistically calculated, and the concentration of the arrival frequency of the encoding task is analyzed. Assume that the absolute value of the frame flow difference exceeding 5 frames is defined as a significant change, and less than or equal to 5 frames is a slight change. The absolute values of the difference in this example are 5 frames, 7 frames, 12 frames and 8 frames respectively, among which changes of 7 frames, 12 frames and 8 frames are defined as significant changes, and changes of 5 frames are defined as slight changes. Based on this, the frequency and concentration of significant changes are recorded, and finally the frequency concentration distribution result is formed. For example, if 75% of the frame flow changes in adjacent time periods reach a significant level, it means that the frequency concentration of the encoding task is high.

[0122] S413: Call the frequency concentration distribution result, filter the task flow that needs to be allocated an exclusive buffer area, and generate the encoding task buffer level;

[0123] Based on the frame stream difference analysis above, the frequency and concentration of significant changes in the number of frame streams are called to screen task flows that require exclusive buffer area allocation. The exclusive buffer allocation threshold is set to a significant change ratio exceeding 60%. In this example, the significant change ratio is 75%, which exceeds the set threshold. Therefore, the video frame input in the current period is determined to be a task flow that requires exclusive buffer area allocation, and its buffer level is marked as high. If the significant change ratio is in the range of [30%, 60%], it is defined as medium level, and if it is below 30%, it is defined as low level. The video frame input sequence buffer level obtained through the above comparative analysis is marked as high level, and the corresponding encoding task buffer level record is generated.

[0124] See also Figure 6 , the specific steps for obtaining the processing channel adjustment record are:

[0125] S511: Call the encoding task buffer level and image structure aggregation results, extract the boundary pixel grayscale information of each image partition in the current frame, record the grayscale change direction and amplitude between adjacent partitions, analyze the grayscale continuous change characteristics at the boundary position, and generate a boundary grayscale change data set;

[0126] Call the encoding task buffer level and image structure aggregation level. Take the currently processed 5th frame image as an example. Call the buffer level and aggregation level information, divide the frame image into 16 areas, and extract the pixel grayscale of each partition boundary in detail. For example, in the boundary area between partition numbers 6 and 7, extract the grayscale values of all adjacent pixel pairs on the boundary line. For example, the grayscale values of the pixel pairs on the boundary line are (120, 125), (125, 130), (130, 135), and (135, 140) respectively. Then record the direction and amplitude of the grayscale difference for each pair. For example, if the grayscale difference of the first pair is 125-120=+5, it represents the direction of grayscale increase, and the amplitude is 5. Similarly, all grayscale change data on the boundary line are obtained and recorded as the grayscale change information of adjacent partition boundaries. The same analysis is completed for all area boundaries to form a complete boundary grayscale change data set. For example, in the partition boundary grayscale change data of the 5th frame, the grayscale difference amplitudes between adjacent areas are [+5, +5, +5, +5], [-3, -2, -4, -3], etc., and all data are stored and recorded as the current frame boundary grayscale change data set.

[0127] S512: Based on the boundary grayscale change dataset, analyze the consistency of the edge grayscale continuity between image partitions, evaluate the stability of the grayscale change amplitude in adjacent regions, identify the transition connection strength of each frame in the spatial distribution, and generate an image coherence score;

[0128] According to the boundary grayscale change data set, taking the adjacent area edges of the 5th frame as an example, the grayscale continuity trend of each pair of partition boundaries is analyzed. Taking the grayscale change data [+5, +5, +5, +5] between areas 6 and 7 as an example, the consistency of the grayscale change direction is judged. If the continuous change direction remains consistent and the difference in the change amplitude is less than the set threshold (the threshold is 2 grayscale levels), it is determined to be a continuous and stable change. For example, if the above data are all in the positive direction and the difference in the change amplitude is 0, which is less than the threshold, the grayscale continuity trend consistency is evaluated as high; the grayscale change of the boundary between adjacent areas 8 and 9 is [-3, -2, -4, -3]. Similarly, the continuity of the grayscale change direction is evaluated and the difference between the change amplitudes is calculated. For example, if the grayscale change amplitude is the largest, The difference is |-2-(-4)|=2, which is just at the threshold boundary and is still considered to have stable grayscale continuity. After completing this analysis for each adjacent partition boundary, the evaluation results are recorded. By calculating the proportion of all boundaries with high grayscale continuity consistency to the total number of boundaries, for example, in this example, there are a total of 24 boundaries, 20 of which show high consistency. The calculated coherence score is 20 / 24=0.833, that is, the image coherence score is 0.833. The coherence score value range is defined as: [0.7, 1] is highly coherent, [0.4, 0.7) is generally coherent, and [0, 0.4) is incoherent. The current frame is scored as highly coherent, and the score result is stored as the image coherence score of the current frame.

[0129] S513: Calling the image coherence score, comparing the connection strength between each partition in the current frame with the grayscale jump distribution of the frame edge area, adjusting the processing channel of each image frame, and establishing a processing channel adjustment record;

[0130] The image coherence score is obtained. Based on the coherence score of 0.833 of the 5th frame, the connection strength between each partition and the grayscale jump distribution of the frame edge area are further analyzed. Taking the connection strength between block numbers 6 and 7 as an example, the connection strength is judged by the grayscale continuity. If the continuous direction of the grayscale change of the boundary exceeds 90% of the pixel pairs, the connection strength is defined as strong; if the grayscale continuous pixel pairs of the boundary between blocks 1 and 2 account for only 50%, the connection strength is defined as weak; at the same time, the grayscale jump characteristics of the overall edge area of ​​the 5th frame (such as the outermost circle boundary of the image) are analyzed. If the number of pixels with grayscale jumps in the edge area accounts for more than 40% of the total edge pixels, it is determined that the grayscale jump in the frame edge area is obvious, and the grayscale jump ratio of the frame edge area is set. It is 45%, which meets the obvious jump judgment; then the connection strength and edge jump distribution are called to dynamically adjust the processing channel of the image frame. For example, for areas with strong connection strength and located inside the image (such as blocks 6 and 7), high-priority processing channels are set to enhance the local detail encoding strength; and for partitions with weak connection strength or located in obvious jump edge areas (such as blocks 1 and 2), they are adjusted to low-priority processing channels to reduce the computational load and avoid excessive encoding of details; the image partition processing channel configuration after channel adjustment is recorded, for example, the channel adjustment marks of blocks 6 and 7 are set as high-priority channels, and blocks 1 and 2 are adjusted to low-priority channels. After completing the adjustment of each area, all channel adjustment configuration records are stored as processing channel adjustment records.

[0131] The video encoding system based on multi-channel concurrent software and hardware hybrid is used to execute the above-mentioned video encoding method based on multi-channel concurrent software and hardware hybrid. The system includes:

[0132] The redundant frame filtering module extracts pixel brightness features, performs edge grayscale structure analysis, and detects regional texture distribution for each frame based on the video frame sequence to be processed. It also analyzes the similarity of each pair of adjacent frames, identifies redundant frames, extracts valid frames, and generates video frame screening records.

[0133] The task sorting module screens and records video frames, analyzes the generation order of valid frames and the dynamic change response characteristics of the corresponding target area, identifies the appearance time of the target frame and the degree of content fluctuation, and calculates the processing priority of each frame by comparing the distribution density of multiple image frames on the time axis, and establishes a task level mapping mark.

[0134] The region combination module divides the image frame into blocks based on the task-level mapping markers, analyzes the degree of pixel brightness change in the same block between adjacent frames, evaluates the continuity of the edge structure between blocks, combines the blocks and constructs independent processing units to establish the image structure aggregation result;

[0135] The buffer allocation module collects video frame traffic within a continuous period based on the image structure aggregation results, analyzes the concentration of encoding task arrival frequency, evaluates the load status in real time, identifies task flows that require exclusive buffer area allocation, and establishes encoding task buffer levels;

[0136] Based on the encoding task buffer level and image structure aggregation results, the channel optimization module detects and records the direction of grayscale change at the partition boundary within each image frame, analyzes the stability of the grayscale transition at the edges between adjacent partitions, evaluates the partition connection strength of the entire frame image, adjusts the processing channel of each image frame, and establishes a processing channel adjustment record.

[0137] The above are merely preferred embodiments of the present invention and do not limit the present invention in any other form. Any technician familiar with the profession may use the technical content disclosed above to change or modify it into an equivalent embodiment with equivalent changes and apply it to other fields. However, any simple modification, equivalent change and modification made to the above embodiment based on the technical essence of the present invention without departing from the content of the technical solution of the present invention shall still fall within the scope of protection of the technical solution of the present invention.

Claims

1. A video encoding method based on multi-channel concurrent software and hardware hybrid, characterized in that: The following steps are involved: S1: Obtain the video frame sequence to be processed, extract the pixel brightness distribution, edge grayscale structure and regional texture contour of each frame image, evaluate the similarity of each pair of adjacent frames, identify redundant frames and filter valid frames, and generate video frame filtering records; S2: Based on the video frame screening records, the generation order of valid frames and the dynamic change response characteristics of the corresponding target area are analyzed, the appearance time and the closeness of the content fluctuation of the target frame are identified, and the processing priority of each frame is calculated by comparing the distribution density of multiple image frames on the time axis to generate a task level mapping mark; S3: calling the task level mapping marker, dividing each frame of the image into multiple blocks, analyzing the degree of pixel brightness change in the same block between adjacent frames, evaluating the continuity of the edge structure between blocks, combining the image blocks into processing units, and generating an image structure aggregation result; S4: Call the image structure aggregation result, obtain the continuous input of the video frame sequence, extract the number of frame flows in a continuous time period, analyze the concentration of the arrival frequency of encoding tasks, evaluate the load status, screen the task flows that need to allocate exclusive buffer areas, and generate the encoding task buffer level.

2. The video encoding method based on multi-channel concurrent software and hardware hybrid according to claim 1, characterized in that: The video frame screening record includes frame continuity judgment results, redundant frame index list, and valid frame sequence; the task level mapping mark includes frame processing level number, priority allocation label, and event trigger sorting result; the image structure aggregation result includes block merging mapping relationship, structural connectivity partition, and block intra-edge structure indication; the encoding task buffer level includes task flow scheduling label, buffer usage ratio information, and frame flow density distribution information.

3. The video encoding method based on multi-channel concurrent software and hardware hybrid according to claim 1, characterized in that: The steps for obtaining the video frame screening record are specifically as follows: S111: Obtain a sequence of video frames to be processed, call pixel distribution data of each frame image, analyze the brightness distribution state, edge grayscale structure trend and texture contour coding mode of each area in the image, and generate basic feature data of the frame image; S112: Based on the basic feature data of the frame images, analyzing the brightness variation range and grayscale structure connection characteristics of adjacent frames in the same area, evaluating the structural stability and texture disturbance intensity between consecutive images, calculating the regional similarity measurement value between adjacent frames, and obtaining the adjacent frame similarity measurement value; S113: calling the adjacent frame similarity metric value, marking redundant frames and adding frame skipping identifiers according to the similarity, screening valid frames and constructing a valid frame sequence, and generating a video frame screening record.

4. The video encoding method based on multi-channel concurrent software and hardware hybrid according to claim 3, characterized in that: The steps for obtaining the task level mapping mark are specifically as follows: S211: Based on the video frame screening record, analyzing the generation order of valid frames in the continuous sequence, obtaining the frame index number and the position relationship between each image frame, and generating a valid frame generation time sequence feature; S212: Calling the valid frame to generate a temporal feature quantity, analyzing the continuous response characteristics of the target area in each frame image, identifying the position change and stay of the target area in the continuous image, extracting the continuous number of the area in the motion path and analyzing the concentration state of the change, and generating a target area response density index; S213: Analyze the central tendency of the image frames in content change and temporal distribution according to the target area response density index, calculate the processing priority score of each frame, construct a processing task sequence, and establish a task level mapping mark.

5. The video encoding method based on multi-channel concurrent software and hardware hybrid according to claim 4, characterized in that: The steps for obtaining the image structure aggregation result are specifically as follows: S311: Based on the task level mapping mark, the current frame is divided into multiple image blocks, and pixel brightness data of corresponding blocks in the current frame and the previous frame are analyzed to identify the degree of pixel brightness change, evaluate the magnitude of the change in image content, and generate a brightness change difference data set; S312: Calling the brightness change difference dataset, analyzing the edge pixel color direction of each pair of adjacent image blocks in the current frame, identifying the consistency of the direction angle, analyzing the continuity of the edge color, and generating edge direction consistency information; S313: Pair and combine image blocks according to the edge direction consistency information, compare the brightness change trend and edge direction consistency between blocks, calculate the merging adaptation coefficient, combine multiple image blocks into independent processing units, and establish an image structure aggregation result.

6. The video encoding method based on multi-channel concurrent software and hardware hybrid according to claim 5, characterized in that: The steps for obtaining the encoding task buffer level are specifically as follows: S411: Calling the image structure aggregation result, obtaining continuous input data of the video frame sequence, analyzing the number of video frames received in each time period, and generating a frame flow statistics record for each time period; S412: Calculating the difference between the number of frame flows in each pair of adjacent time periods based on the frame flow statistics for the time period, identifying the continuity of the direction of change in the number of frame flows, analyzing the concentration of the arrival frequency of the encoding task, and generating a frequency concentration distribution result; S413: Call the frequency concentration distribution result, filter the task flow that needs to be allocated an exclusive buffer area, and generate the encoding task buffer level.

7. The video encoding method based on multi-channel concurrent software and hardware hybrid according to claim 1, characterized in that: The method further comprises: S5: Calling the encoding task buffer level and image structure aggregation results, extracting boundary grayscale change information of multiple partitions in the current frame, evaluating the consistency and coherence of the overall image transition by analyzing the edge continuity and grayscale stability between partitions, comparing the connection strength of internal partitions of the image and the external jump distribution frequency, adjusting the processing channel of each image frame, and generating a processing channel adjustment record; The processing channel adjustment record includes channel allocation instructions, inter-frame channel change sequence, and channel load association mapping information.

8. The video encoding method based on multi-channel concurrent software and hardware hybrid according to claim 7, characterized in that: The specific steps for obtaining the processing channel adjustment record are: S511: Calling the encoding task buffer level and image structure aggregation result, extracting the boundary pixel grayscale information of each image partition in the current frame, recording the grayscale change direction and amplitude between adjacent partitions, analyzing the grayscale continuous change characteristics at the boundary position, and generating a boundary grayscale change data set; S512: Analyze the consistency of edge grayscale continuity between image partitions based on the boundary grayscale change dataset, evaluate the stability of grayscale change amplitudes in adjacent regions, identify the transition connection strength of each frame in spatial distribution, and generate an image coherence score; S513: Calling the image coherence score, comparing the connection strength between each partition in the current frame with the grayscale jump distribution of the frame edge area, adjusting the processing channel of each image frame, and establishing a processing channel adjustment record.

9. A video coding system based on multi-channel concurrent software and hardware hybrid, characterized in that: The system is used to implement the video encoding method based on multi-channel concurrent software and hardware hybrid according to any one of claims 1 to 8, and the system includes: The redundant frame filtering module extracts pixel brightness features, performs edge grayscale structure analysis, and detects regional texture distribution for each frame based on the video frame sequence to be processed. It also analyzes the similarity of each pair of adjacent frames, identifies redundant frames, extracts valid frames, and generates video frame screening records. The task sorting module screens the video frames and analyzes the generation order of valid frames and the dynamic change response characteristics of the corresponding target area based on the video frame screening record, identifies the appearance time of the target frame and the closeness of the content fluctuation, calculates the processing priority of each frame by comparing the distribution density of multiple image frames on the time axis, and establishes a task level mapping mark; The region combination module divides the image frame into blocks based on the task level mapping mark, analyzes the degree of pixel brightness change in the same block between adjacent frames, evaluates the continuity of the edge structure between blocks, combines the blocks and constructs independent processing units to establish the image structure aggregation result; The buffer allocation module collects the video frame flow in a continuous period based on the image structure aggregation result, analyzes the concentration of the encoding task arrival frequency, evaluates the load status in real time, identifies the task flow that needs to be allocated an exclusive buffer area, and establishes the encoding task buffer level; Based on the encoding task buffer level and image structure aggregation results, the channel optimization module detects and records the direction of grayscale change at the partition boundary within each image frame, analyzes the stability of the grayscale transition at the edges between adjacent partitions, evaluates the partition connection strength of the entire frame image, adjusts the processing channel of each image frame, and establishes a processing channel adjustment record.

Images