A high-definition video transmission method and system based on diversified data compression

Abstract

This invention discloses a high-definition video transmission method and system based on diversified data compression. The method includes the following steps: a diversified data compression system standardizes video images and calculates scene coefficients; an adaptive monitoring interval is established based on the scene coefficients, and an inter-frame estimation algorithm is designed within the adaptive monitoring interval to obtain an inter-frame video image set; an intra-frame prediction algorithm is designed for the inter-frame video image set to obtain an intra-frame compressed image set; adjacent video images are randomly selected from the intra-frame compressed image set, and a comprehensive compression evaluation algorithm is designed to obtain a compressed video image set; a subset of similar video images is extracted from the compressed video image set, and the subset of similar video images is recombined to obtain a video output subset, which is then transmitted as a whole. This scheme significantly improves the compression efficiency of high-definition video by designing intra-frame compression and inter-frame redundant information removal mechanisms, effectively improving the compression efficiency and real-time transmission performance of video images.

Description

Technical Field

[0001] This invention belongs to the field of video image compression technology, and in particular relates to a high-definition video transmission method and system based on diversified data compression. Background Technology

[0002] With the rapid development of ultra-high-definition video technology, high-definition video is increasingly widely used in fields such as remote monitoring and intelligent transportation. However, high-definition video has inherent characteristics such as high resolution, high frame rate, and large data volume. The data volume of a single frame of 4K video image can reach tens of megabytes, requiring a huge amount of network bandwidth resources for continuous transmission. In the current network environment, the contradiction between massive high-definition video data and limited transmission bandwidth is becoming increasingly prominent, becoming a core bottleneck restricting the large-scale development of the high-definition video industry. To alleviate this contradiction, video compression technology has become a key support in the high-definition video transmission system. Traditional compression technologies mostly adopt single inter-frame or intra-frame compression strategies, lacking precise adaptation to dynamic changes in video scenes and internal structural features of images, and cannot specifically mine intra-frame redundancy, resulting in low compression efficiency. In the inter-frame redundancy screening stage, fixed parameter settings are prone to causing redundant residues or loss of effective information, making it difficult to balance compression efficiency and image quality preservation.

[0003] Meanwhile, existing compression technologies have significant shortcomings in their compression quality evaluation systems and transmission optimization designs. Most compression schemes use a single metric to evaluate compression effectiveness, which cannot comprehensively reflect the actual image quality and transmission adaptability after video compression. This easily leads to problems such as over-compression causing image quality degradation or under-compression causing bandwidth waste. Furthermore, existing technologies do not pay enough attention to the recombination and optimization of similar video data after compression. The repeated transmission of a large number of similar frames further exacerbates the bandwidth burden and reduces transmission efficiency. Therefore, there is an urgent need to propose a new video image compression method. Summary of the Invention

[0004] This invention provides a high-definition video transmission method and system based on diversified data compression to solve the technical problems of redundant video image data, low compression efficiency, and lack of compression monitoring.

[0005] To solve the above-mentioned technical problems, the present invention provides the following technical solution: On one hand, the present invention provides a high-definition video transmission method based on diversified data compression, the high-definition video transmission method based on diversified data compression includes: S1: Standardize the video image, use Gaussian filtering to remove noise from the video image, and perform grayscale processing. Design a large-scale feature extraction algorithm based on the grayscale of the video image to calculate the scene coefficient. S2: Establish an adaptive monitoring interval based on scene coefficients, design an inter-frame estimation algorithm within the adaptive monitoring interval, calculate the inter-frame estimation coefficients of adjacent video images, initialize the dynamic estimation threshold, and delete redundant video images based on the inter-frame estimation coefficients and the dynamic estimation threshold to obtain an inter-frame video image set. S3: Design an intra-frame prediction algorithm for inter-frame video image sets. Calculate the classification coefficient and the clustering coefficient of each classification coefficient using the texture features of the video images. Compress and reassemble the large clusters in the video images to obtain an intra-frame compressed image set. S4: Randomly select adjacent video images from the intra-frame compressed image set, design a comprehensive compression evaluation algorithm, calculate the intra-frame compression ratio, initialize the compression ratio threshold, if the intra-frame compression ratio is less than the compression ratio threshold, perform compensation compression, if the intra-frame compression ratio is greater than or equal to the compression ratio threshold, do not perform compensation compression, and obtain a compressed video image set. S5: Extract a subset of similar video images from the compressed video image set, reassemble the subset of similar video images to obtain the video output subset, and transmit the video output subset as a whole.

[0006] Preferably, step S1 includes: The video images are standardized by converting the raw video image data into YUV420 format and uniformly controlling the frame rate. Missing frames are then supplemented by inter-frame interpolation. Gaussian filtering is used to remove noise from the video image; the video image after Gaussian filtering is then converted to grayscale, and the grayscale values ​​of the video image are extracted. Each pixel contains a grayscale value, resulting in a grayscale image. Design a large-scale feature extraction algorithm based on video image grayscale and calculate scene coefficients. The specific steps are as follows: (1) Gray-scale layering and large-scale block division: The gray-scale image is divided into 5 intervals [0~50], [51~100], [101~150], [151~200], and [201~255] according to the gray-scale value, and large-scale non-overlapping blocks are performed with a block size of 1 / 32 of the video resolution to obtain multiple large-scale blocks; (2) Block gray-scale clustering and statistics: perform cluster analysis on the gray-scale interval distribution in each large-scale block, determine the cluster center, take the peak pixel proportion of each gray-scale interval as the cluster core, and count the number of clusters in each large-scale block and the pixel proportion corresponding to the largest cluster. (3) Scene coefficient fusion calculation: The number of clusters of all large-scale blocks is averaged to obtain the average number of clusters C. The specific calculation formula of the final scene coefficient S is as follows: ; Where Cmax is the preset maximum cluster size threshold, Rmax is 100%, S ranges from [0,1], and R represents the average percentage. For open scenes, S≥0.7, and for dense scenes, S≤0.4. and These represent the proportion coefficients of the average percentage and the average number of clusters, respectively.

[0007] Preferably, step S2 includes: An adaptive monitoring interval is established based on the scene coefficient, which has a value range of [0,1]. The video image processed in step S1 is divided into n interval levels. Different interval levels correspond to different monitoring interval lengths. The number of all monitoring interval lengths in the video image is counted, and the mode of the monitoring interval length is used as the final adaptive monitoring interval length. Within the adaptive monitoring interval, an inter-frame estimation algorithm is designed to calculate the sum of the absolute values ​​of the pixel grayscale differences between the frame to be detected and the reference frame, where the frame to be detected is the frame following the next frame between two adjacent frames, and the reference frame is the frame preceding the next frame between two adjacent frames. The similarity between the two frames is quantified, and the specific formula is as follows: + ; Where D represents the sum of the absolute values ​​of the pixel grayscale differences. Let be the pixel grayscale value of the frame to be detected at position (i,j). Let be the pixel grayscale value at position (i,j) of the reference frame, and M and N be the height and width of the image, respectively. This represents the deviation of the cluster center coordinates between the frame to be detected and the reference frame, where || represents the absolute value calculation. The sum of the absolute values ​​of the pixel grayscale differences is normalized, and the inter-frame estimation coefficients of adjacent video images are calculated. The range of the inter-frame estimation coefficients is [0,1]. The closer the coefficient is to 0, the higher the similarity between the frame to be detected and the reference frame, and the greater the redundancy. The closer the coefficient is to 1, the lower the similarity between the frame to be detected and the reference frame, and the smaller the redundancy. The dynamic estimation threshold is initialized based on the scene complexity of the video image. The calculated inter-frame estimation coefficients are compared with the dynamic estimation threshold. If the inter-frame estimation coefficients are less than or equal to the threshold, the frame to be detected is determined to be a redundant frame and deleted. If the inter-frame estimation coefficients are greater than the threshold, the frame to be detected is determined to be a valid frame and retained. The process involves comparing all video images within the adaptive monitoring interval, identifying and deleting redundant frames, and obtaining an inter-frame video image set.

[0008] Preferably, the deviation value of the cluster center coordinates between the frame to be detected and the reference frame is calculated based on the coordinates of the center points between the clusters, and the specific steps are as follows: (1) Use the DBSCAN algorithm to calculate the number of clusters of the frame to be detected and the number of pixels contained in each cluster. Use the DBSCAN algorithm to calculate the number of clusters of the reference frame and the number of pixels contained in each cluster. (2) Set the cluster set of the frame to be detected and the cluster set of the reference frame. Use the inter-frame matching algorithm to calculate the minimum value of the cluster center point coordinates in the cluster set of the frame to be detected and the cluster set of the reference frame. The minimum value is the deviation value of the cluster center point coordinates between the frame to be detected and the reference frame.

[0009] Preferably, step S3 includes: (1) Perform grayscale gradient calculation on single-frame images in the inter-frame video image set, use the Sobel operator to extract the horizontal and vertical gradients of the image, calculate the texture direction and texture intensity of the pixel through the gradient value, and divide the image into three types of regions based on the texture direction and texture intensity: smooth region, edge region, and complex texture region. (2) Calculate the classification coefficient and clustering coefficient. The classification coefficient is the proportion of each region in the image. The K-Medoids clustering algorithm is used to calculate the clustering coefficient, which is the number of pixels contained in each cluster. Initialize the large clustering threshold. For clusters with a clustering coefficient greater than the large clustering threshold, they are marked as large clusters. (3) Use the classification coefficient to compress and reorganize the large clusters to obtain the intra-frame compressed set.

[0010] On the other hand, the present invention also provides a high-definition video transmission system based on diversified data compression, the high-definition video transmission system based on diversified data compression comprising: The preprocessing module standardizes the video images, removes noise from the video images using Gaussian filtering, performs grayscale processing, designs a large-scale feature extraction algorithm based on the grayscale of the video images, and calculates scene coefficients. The inter-frame compression module establishes an adaptive monitoring interval based on scene coefficients, designs an inter-frame estimation algorithm within the adaptive monitoring interval, calculates the inter-frame estimation coefficients of adjacent video images, initializes a dynamic estimation threshold, and removes redundant video images based on the inter-frame estimation coefficients and the dynamic estimation threshold to obtain an inter-frame video image set. The intra-frame compression module designs an intra-frame prediction algorithm for the inter-frame video image set. It uses the texture features of the video images to calculate the classification coefficient and the clustering coefficient of each classification coefficient, and compresses and reassembles the large clusters in the video images to obtain the intra-frame compressed image set. The compensation compression and output module randomly selects adjacent video images from the intra-frame compressed image set, designs a comprehensive compression evaluation algorithm, calculates the intra-frame compression ratio, initializes a compression ratio threshold, and performs compensation compression if the intra-frame compression ratio is less than the compression ratio threshold; otherwise, it does not perform compensation compression, resulting in a compressed video image set. It then extracts a subset of similar video images from the compressed video image set, reassembles the similar video image subsets to obtain a video output subset, and transmits the entire video output subset.

[0011] The beneficial effects of the technical solution provided by this invention include at least the following: 1. This invention significantly improves the compression efficiency of high-definition video through targeted intra-frame compression design and a full-process redundancy removal mechanism, effectively alleviating the contradiction between massive video data and limited transmission bandwidth. Compared to traditional single compression strategies, this invention innovatively combines video image texture features in the intra-frame compression stage to calculate classification and clustering coefficients, and performs targeted compression and recombination on large clusters. Traditional intra-frame compression often uses uniform compression, which cannot fully exploit redundancy in images containing large, similar targets. This invention, however, accurately locates large clusters through cluster analysis and employs targeted compression algorithms to maximize the removal of intra-frame redundancy while ensuring the integrity of the cluster structure, improving intra-frame compression efficiency by more than 40%. Simultaneously, an adaptive monitoring interval is constructed based on scene features, and redundant images are accurately screened through inter-frame estimation algorithms. Similar video image subsets are extracted and recombined for transmission, ensuring the continuity of video transmission and further improving the accuracy of the compressed video quality.

[0012] 2. This invention, through precise control across multiple stages and a scientific compression evaluation system, improves compression efficiency while preserving the image quality details of high-definition video to the maximum extent, significantly enhancing the user viewing experience. Traditional compression technologies often suffer from image quality degradation issues such as blurred edges, lost textures, and distorted object contours due to a lack of targeted compression strategies or a single evaluation metric. This invention targets large clusters for compression and recombination. Through precise calculation of classification and clustering coefficients, it can accurately identify and preserve the key structural and texture information of large clusters. By designing a comprehensive compression evaluation algorithm, combined with a compression rate threshold, it achieves flexible control of compensatory compression. When the compression rate is below the threshold, compensatory compression prevents over-compression and image quality degradation; when the compression rate reaches the target, compensation stops, avoiding resource waste caused by over-processing, perfectly balancing the dual requirements of compression efficiency and image quality preservation.

[0013] 3. This invention possesses strong scene adaptability and transmission stability, enabling it to be widely adapted to various high-definition video application scenarios and complex network environments, demonstrating excellent versatility and practicality. Through a fully adaptive design, this invention achieves precise adaptation to diverse scenarios: it dynamically adjusts inter-frame monitoring intervals and estimated thresholds based on scene features, calculates classification and clustering coefficients for images with different texture features, and achieves efficient compression for images containing different types of large clusters. Simultaneously, the similar subset recombination design not only reduces data volume but also enhances the anti-interference capability of data transmission. Combined with the overall transmission strategy, it effectively addresses network packet loss and latency issues, improving transmission stability by over 50%. Furthermore, the strong independence and good synergy of each module in this invention facilitate engineering implementation and subsequent technological iteration, allowing for rapid deployment in multiple fields such as remote monitoring, video conferencing, and ultra-high-definition video-on-demand, demonstrating broad market application prospects and significant economic and social benefits. Attached Figure Description

[0014] Figure 1 A schematic diagram of the overall execution flow of a high-definition video transmission method based on diversified data compression provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of large-scale block partitioning for a high-definition video transmission method based on diversified data compression, provided as an embodiment of the present invention. Detailed Implementation

[0015] The present invention will be further described below with reference to the accompanying drawings, but this is not intended to limit the present invention in any way. Any modifications or substitutions made based on the teachings of the present invention shall fall within the protection scope of the present invention. Example 1

[0016] This embodiment provides a high-definition video transmission method based on diversified data compression, which can be implemented by electronic devices, such as... Figure 1 As shown. Specifically, the method in this embodiment includes the following steps: To ensure the fault tolerance of this solution and expand its application scope, the video images are standardized by converting the acquired raw video image data into YUV420 format. This ensures that all video images are processed under a unified standard format, improving the accuracy and efficiency of video image compression. The frame rate is also standardized to 30fps, where fps stands for frames per second. Missing frames are supplemented through frame interpolation. Gaussian filtering is used to remove noise from video images. A filter kernel is constructed based on the normal distribution function. Noise smoothing is achieved by weighted averaging of gray values ​​in the neighborhood of a pixel, which effectively suppresses Gaussian noise while preserving edge details of the video image to the greatest extent. The video image after Gaussian filtering is converted to grayscale, and the grayscale values ​​of the video image are extracted. Each pixel contains a grayscale value, and a grayscale image is generated. Design a large-scale feature extraction algorithm based on video image grayscale and calculate scene coefficients. The specific steps are as follows: (1) Gray-scale layering and large-scale segmentation: The gray-scale image is divided into 5 intervals according to gray-scale values ​​[0~50], [51~100], [101~150], [151~200], and [201~255], and large-scale non-overlapping segments are performed with a block size of 1 / 32 of the video resolution to obtain multiple large-scale blocks; such as Figure 2 As shown, the entire video frame is divided into large-scale blocks. Under normal circumstances, different large-scale blocks do not overlap, thus preserving the features of large-scale regions while reducing computational complexity. (2) Block gray-scale clustering and statistics: cluster analysis is performed on the distribution of gray-scale intervals in each large-scale block. The distance maximization method is used to determine the cluster center. The peak pixel proportion of each gray-scale interval is used as the cluster core. The number of clusters in each large-scale block and the pixel proportion corresponding to the largest cluster are statistically analyzed. In urban dense scenes, due to the large number of targets such as buildings and people, the number of clusters in the block is large and the proportion of the largest cluster is dispersed. In open scenes, due to the single target, the number of clusters in the sky and the ground is small and the proportion of the largest cluster is concentrated. (3) Scene coefficient fusion calculation: The number of clusters of all large-scale blocks is averaged to obtain the average number of clusters C. The proportion of the largest cluster pixels of all large-scale blocks is weighted and averaged to obtain the average proportion R. The edge block weight is 0.8 and the center block weight is 1.2. The feature weight of the core area is increased. The specific calculation formula of the final scene coefficient S is as follows: ; Where Cmax is the preset maximum cluster size threshold, Rmax is 100%, and S ranges from [0,1], with S≥0.7 for open scenes and S≤0.4 for dense scenes. and These represent the proportion coefficients of the mean percentage and the mean number of clusters, respectively. It should be further noted that... and These are preset values, set based on historical experience, and + .

[0017] After determining the scene coefficients, a preliminary judgment of the video image can be made, and different compression strategies can be established based on the scene coefficients. An adaptive monitoring interval can be established based on the scene coefficients. Since the value range of the scene coefficients is [0,1], the processed video image can be divided into n interval levels. Different interval levels correspond to different monitoring interval lengths. The value of n is generally 5, that is, the 5 interval levels are [0,0.2], [0.2,0.4], [0.4,0.7], [0.7,0.8], [0.8,1]. The number of all monitoring interval lengths in the video image is counted, and the monitoring interval length with the most counts is taken as the final adaptive monitoring interval length. For image blocks with small scene coefficients, the video frames change less and the inter-frame redundancy is high. A longer monitoring interval can cover more redundant frames. In other words, for simple static scenes, a longer monitoring interval length should be set. For image blocks with large scene coefficients, the video frames change rapidly and have low inter-frame redundancy. A shorter monitoring interval can avoid the loss of effective frames due to an excessively long monitoring interval. In other words, for complex dynamic scenes, a shorter monitoring interval length should be set. The establishment of adaptive monitoring intervals enables dynamic matching between the length of the monitoring interval and the complexity of the scene, which significantly improves the accuracy of inter-frame redundancy detection compared to traditional fixed-length monitoring intervals. It should be further explained that in actual operation, if the video image is an open scene, the adaptive monitoring interval length is 1000, which corresponds to 1000 frames of video image, corresponding to the level interval [0.8,1]. After that, each level decreases by 100. If the video image is a dense scene, the adaptive monitoring interval length is 400, which corresponds to 400 frames of video image, corresponding to the level interval [0.8,1]. After that, each level decreases by 50. Within the adaptive monitoring interval, an inter-frame estimation algorithm is designed to calculate the sum of the absolute values ​​of the pixel grayscale differences between the frame to be detected and the reference frame, where the frame to be detected is the frame following the next frame between two adjacent frames, and the reference frame is the frame preceding the next frame between two adjacent frames. The similarity between the two frames is quantified, and the specific formula is as follows: + ; Where D represents the sum of the absolute values ​​of the pixel grayscale differences. Let be the pixel grayscale value of the frame to be detected at position (i,j). Let be the pixel grayscale value at position (i,j) of the reference frame, and M and N be the height and width of the image, respectively. This represents the deviation of the cluster center coordinates between the frame to be detected and the reference frame; The sum of the absolute values ​​of pixel grayscale differences is normalized to calculate the inter-frame estimation coefficients of adjacent video images. Since the sum of the absolute values ​​of pixel grayscale differences is too large, it will affect the calculation efficiency and accuracy. Therefore, the sum of the absolute values ​​of pixel grayscale differences can be uniformly reduced so that the value range of the inter-frame estimation coefficients is [0,1]. For example, dividing the sum of the absolute values ​​of pixel grayscale differences by 100, the closer the coefficient is to 0, the higher the similarity between the frame to be detected and the reference frame, and the greater the redundancy; the closer the coefficient is to 1, the lower the similarity between the frame to be detected and the reference frame, and the smaller the redundancy. It should be further explained that, after calculation, all video images except the first frame will obtain the sum of the absolute values ​​of the pixel grayscale differences; The dynamic estimation threshold is initialized based on the scene complexity of the video image: for simple static scenes, a lower estimation threshold is set to ensure that more redundant frames are identified and deleted; for complex dynamic scenes, a higher estimation threshold is set to avoid the deletion of valid frames. The calculated inter-frame estimation coefficients are compared with the dynamic estimation threshold. If the inter-frame estimation coefficients are less than or equal to the threshold, the frame to be detected is determined to be a redundant frame and deleted; if the inter-frame estimation coefficients are greater than the threshold, the frame to be detected is determined to be a valid frame and retained. The process continues until all video images within the adaptive monitoring interval are compared, allowing more redundant frames to be identified and removed, resulting in a set of inter-frame video images.

[0018] The deviation of the cluster center coordinates between the frame to be detected and the reference frame is calculated based on the coordinates of the center points between the clusters. The specific steps are as follows: (1) Use the DBSCAN algorithm to calculate the number of clusters of the frame to be detected and the number of pixels contained in each cluster. Use the DBSCAN algorithm to calculate the number of clusters of the reference frame and the number of pixels contained in each cluster. (2) Set the cluster set of the frame to be detected, and sort the number of pixels in each cluster of the cluster set of the frame to be detected from small to large. Set the cluster set of the reference frame, and sort the number of pixels in each cluster of the cluster set of the reference frame from small to large. (3) After sorting the cluster set of the frame to be detected and the cluster set of the reference frame, the inter-frame matching algorithm can first perform matching on the sorted set, and then use the inter-frame matching algorithm to perform random matching to calculate the minimum value of the coordinates of the cluster center points in the cluster set of the frame to be detected and the cluster set of the reference frame. The calculation of the minimum value is to calculate the sum of the absolute values ​​of the row and column differences of each cluster center point. When different cluster centers are matched with each other, a minimum value will appear. The above minimum value is the deviation value of the cluster center coordinate points between the frame to be detected and the reference frame. According to this scheme, the minimum value is calculated after sorting and corresponding matching. In order to improve real-time performance and maintain the random matching scheme, after corresponding matching of the coordinates of the cluster center points in the cluster set of the frame to be detected and the cluster set of the reference frame, a random scheme is adopted. The random number is designed to be half the number of cluster center points, following the rounding principle. For example, if the cluster set of the frame to be detected is { , , , The cluster set of the reference frame is { , , , }, calculate the coordinates of the cluster center points when the coordinates of the cluster center points in the cluster set of the frame to be detected and the cluster set of the reference frame match, that is, and , and , and , and The absolute value of the difference yields the result. Then, a random matching scheme was used to obtain the results. and Normally, a threshold needs to be set, but since sorting and matching have been implemented, the minimum value obtained within a limited number of attempts, while taking into account computing power, is the theoretical minimum value.

[0019] An intra-frame prediction algorithm is designed for inter-frame video image sets. It utilizes the texture features of the video images to calculate classification coefficients and clustering coefficients for each classification coefficient. Large clusters in the video images are then compressed and recombined to obtain an intra-frame compressed image set. The specific steps are as follows: (1) Perform grayscale gradient calculation on single-frame images in the inter-frame video image set, use the Sobel operator to extract the horizontal and vertical gradients of the image, calculate the texture direction and texture intensity of the pixel through the gradient value, and divide the image into three types of regions based on the texture direction and texture intensity: smooth region, edge region, and complex texture region. (2) Calculate the classification coefficient and clustering coefficient. Classification coefficient The formula used to quantify the proportion of each region in an image is shown below: ; in, This represents the number of pixels in the u-th class region, where u takes the value [1,3]. When u is 1, it corresponds to a smooth region; when u is 2, it corresponds to an edge region; and when u is 3, it corresponds to a complex texture region. The clustering coefficient is calculated for macroblocks within each class region using the K-Medoids clustering algorithm. The clustering coefficient is the number of pixels contained in each cluster. Large clusters are defined as clusters with large clustering coefficients. Specifically, a large clustering threshold is initialized, and clusters with clustering coefficients greater than the large clustering threshold are marked as large clusters. (3) The large clusters are compressed and recombined using the classification coefficients to obtain an intra-frame compressed image set. For the large clusters, a compression method combining run-length encoding (RLE) and Huffman coding is adopted. For the macroblock gray value sequence within the same classification coefficient, the continuous repeated gray values ​​are first compressed by run-length encoding, and the combination of "gray value - number of repetitions" is used to replace the original sequence. Then, the result of run-length encoding is Huffman encoded, and the optimal Huffman tree is constructed according to the probability of occurrence of gray value and number of repetitions to achieve lossless compression. For small clusters with small clustering coefficients, conventional intra-frame coding is adopted. After compression and recombination, an intra-frame compressed image set is obtained.

[0020] For the obtained intra-frame compressed image set, a compression ratio feedback mechanism is constructed. The compression effect is quantified by a comprehensive compression evaluation algorithm. Whether to perform compensation compression is determined based on the compression ratio threshold, thereby achieving closed-loop optimization of the compression effect. This is different from the one-way processing mode without feedback of traditional compression algorithms, and improves the reliability and adaptability of the compression system. A comprehensive compression evaluation algorithm is constructed by randomly selecting adjacent video images from the intra-frame compressed image set. To ensure the objectivity and representativeness of the evaluation results, a random sampling method is used to select 10 consecutive frames from the intra-frame compressed image set as evaluation samples. The core metric for comprehensive compression evaluation is the intra-frame compression ratio, which is the amount of data in a single frame before compression. In addition to the compression ratio, Peak Signal-to-Noise Ratio (PSNR) is introduced as an image quality evaluation metric, which represents the width and height of the image, respectively. The comprehensive compression evaluation coefficient is obtained through weighted calculation. Then, the compression ratio threshold is initialized and it is determined whether to perform compensation compression. If the intra-frame compression ratio is less than the compression ratio threshold, compensation compression is performed. If the intra-frame compression ratio is greater than or equal to the compression ratio threshold, compensation compression is not performed, resulting in a set of compressed video images.

[0021] Specifically, by testing and analyzing a large number of high-definition video compression samples, the optimal threshold for intra-frame compression ratio was determined. This threshold indicates that when the compression ratio reaches 65%, it can meet the bandwidth requirements of high-definition video transmission while ensuring the subjective acceptability of the image quality, indicating that the current compression effect meets the requirements and no compensation compression is needed; otherwise, compensation compression will be performed. Compensated compression employs a layered quantization optimization strategy, adjusting the quantization step size for different regions of the intra-frame compressed image: setting the baseline compensation to... For smooth regions, increase the quantization step size, setting the quantization step size to [value missing]. This further reduces the amount of data; for edge regions and complex texture regions, the quantization step size remains unchanged to avoid image quality loss. The specific implementation of layered quantization involves multiplying the quantized values ​​in the original quantization matrix by a compensation coefficient. ; A compressed video image set is obtained through compensatory compression optimization. During the compensatory compression process, the changes in compression ratio and PSNR are monitored in real time. When the compression ratio reaches a threshold, compensatory compression is stopped immediately to avoid image quality degradation caused by over-compression. It should be further noted that this solution ensures that the compressed video image meets both the transmission bandwidth requirements and provides a good subjective viewing experience.

[0022] Extract a subset of similar video images from the compressed video image set, initialize a similarity threshold, and if the absolute value of the difference between the inter-frame estimation coefficients of two adjacent video images is less than or equal to the similarity threshold, the adjacent video images are similar and are added to the same subset of similar video images. If the absolute value of the difference between the inter-frame estimation coefficients of two adjacent video images is greater than a similarity threshold, then the adjacent video images are similar, and they are added to a subset of similar video images. For example, the set {( , , , ),( , ),( , )},in( , , , ), ( , ), ( , ) represents different subsets of similar video images, and Z represents video images. This represents the first video image in the compressed video image set; Reorganize similar video image subsets. If the number of similar video image subsets is odd, take the middle video image as the boundary video image. If the number of similar video image subsets is even, take the first of the two middle video images as the boundary video image and incorporate the boundary video image into the previous video output subset. By recombining similar video image subsets using boundary video images, a video output subset is obtained, such as ( , , , The boundary video image in ) is , ( , The boundary video image in ) is The recombined video output subset ( , , ), ( , ), ( , , ), ( ); Transmitting the entire video output subset as a whole can ensure the continuity of video quality while maintaining the compression rate; Example 2

[0023] This embodiment provides a high-definition video transmission system based on diversified data compression, which includes the following modules: The preprocessing module uses a diversified data compression system to standardize video images, employs Gaussian filtering to remove noise from video images, designs a large-scale feature extraction algorithm based on video image grayscale, and calculates scene coefficients. The inter-frame compression module establishes an adaptive monitoring interval based on scene coefficients, designs an inter-frame estimation algorithm within the monitoring interval, calculates the inter-frame estimation coefficients of adjacent video images, initializes the estimation threshold, and removes redundant video images from the inter-frame estimation coefficients and the estimation threshold to obtain an inter-frame video image set. The intra-frame compression module designs an intra-frame prediction algorithm for the inter-frame video image set. It uses the texture features of the video images to calculate the classification coefficient and the clustering coefficient of each classification coefficient, and compresses and reassembles the large clusters in the video images to obtain the intra-frame compressed image set. The compensation compression and output module randomly selects adjacent video images from the intra-frame compressed image set, designs a comprehensive compression evaluation algorithm, calculates the intra-frame compression ratio, initializes a compression ratio threshold, performs compensation compression if the intra-frame compression ratio is less than the compression ratio threshold, and does not perform compensation compression if the intra-frame compression ratio is greater than or equal to the compression ratio threshold, thus obtaining a compressed video image set; extracts a subset of similar video images from the compressed video image set, reassembles the similar video image subsets to obtain a video output subset, and transmits the entire video output subset; As used herein, the term "preferred" is meant as an example, illustration, or illustration. Any aspect or design described herein as "preferred" need not be construed as being more advantageous than other aspects or designs. Rather, the use of the term "preferred" is intended to present the concept in a specific manner. As used in this application, the term "or" is intended to mean an inclusive "or" rather than an exclusionary "or." That is, unless otherwise specified or clear from the context, "X uses A or B" naturally includes either of the permutations. That is, if X uses A; X uses B; or X uses both A and B, then "X uses A or B" is satisfied in any of the foregoing examples.

[0024] Furthermore, although this disclosure has been shown and described with respect to one or more implementations, equivalent variations and modifications will occur to those skilled in the art based on a reading and understanding of this specification and the accompanying drawings. This disclosure includes all such modifications and variations and is limited only by the scope of the appended claims. In particular, with respect to the various functions performed by the aforementioned components (e.g., elements, etc.), the terminology used to describe such components is intended to correspond to any component (unless otherwise indicated) that performs the specified function of said component (e.g., is functionally equivalent to it), even if structurally not equivalent to the disclosed structure performing the functions in the exemplary implementations of this disclosure shown herein. Moreover, although specific features of this disclosure have been disclosed with respect to only one of several implementations, such features may be combined with one or more features of other implementations that may be desirable and advantageous for a given or particular application. Furthermore, with regard to the use of the terms “comprising,” “having,” “containing,” or variations thereof in the Detailed Description or claims, such terms are intended to be included in a manner similar to the term “including.”

[0025] The functional units in this invention embodiment can be integrated into a processing module, or each unit can exist physically separately, or multiple units can be integrated into a module. The integrated module can be implemented in hardware or as a software functional module. If the integrated module is implemented as a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. The storage medium mentioned above can be a read-only memory, a disk, or an optical disk, etc. The aforementioned devices or systems can execute the storage methods in the corresponding method embodiments.

[0026] In summary, the above embodiments are one implementation of the present invention, but the implementation of the present invention is not limited to the embodiments described above. Any changes, modifications, substitutions, combinations, or simplifications made that deviate from the spirit and principle of the present invention should be considered equivalent substitutions and are included within the protection scope of the present invention.

Claims

1. A high-definition video transmission method based on diversified data compression, characterized in that, Includes the following steps: S1: Standardize the video image, use Gaussian filtering to remove noise from the video image, and perform grayscale processing. Design a large-scale feature extraction algorithm based on the grayscale of the video image to calculate the scene coefficient. S2: Establish an adaptive monitoring interval based on scene coefficients, design an inter-frame estimation algorithm within the adaptive monitoring interval, calculate the inter-frame estimation coefficients of adjacent video images, initialize the dynamic estimation threshold, and delete redundant video images based on the inter-frame estimation coefficients and the dynamic estimation threshold to obtain an inter-frame video image set. S3: Design an intra-frame prediction algorithm for inter-frame video image sets. Calculate the classification coefficient and the clustering coefficient of each classification coefficient using the texture features of the video images. Compress and reassemble the large clusters in the video images to obtain an intra-frame compressed image set. S4: Randomly select adjacent video images from the intra-frame compressed image set, design a comprehensive compression evaluation algorithm, calculate the intra-frame compression ratio, initialize the compression ratio threshold, if the intra-frame compression ratio is less than the compression ratio threshold, perform compensation compression, if the intra-frame compression ratio is greater than or equal to the compression ratio threshold, do not perform compensation compression, and obtain a compressed video image set. S5: Extract a subset of similar video images from the compressed video image set, reassemble the subset of similar video images to obtain the video output subset, and transmit the video output subset as a whole.

2. The high-definition video transmission method based on diversified data compression according to claim 1, characterized in that, Step S1 includes: The video images are standardized by converting the raw video image data into YUV420 format and uniformly controlling the frame rate. Missing frames are then supplemented by inter-frame interpolation. Gaussian filtering is used to remove noise from the video image; the video image after Gaussian filtering is then converted to grayscale, and the grayscale values ​​of the video image are extracted. Each pixel contains a grayscale value, resulting in a grayscale image. Design a large-scale feature extraction algorithm based on video image grayscale and calculate scene coefficients. The specific steps are as follows: (1) Gray-scale layering and large-scale block division: The gray-scale image is divided into 5 intervals [0~50], [51~100], [101~150], [151~200], and [201~255] according to the gray-scale value, and large-scale non-overlapping blocks are performed with a block size of 1 / 32 of the video resolution to obtain multiple large-scale blocks; (2) Block gray-scale clustering and statistics: perform cluster analysis on the gray-scale interval distribution in each large-scale block, determine the cluster center, take the peak pixel proportion of each gray-scale interval as the cluster core, and count the number of clusters in each large-scale block and the pixel proportion corresponding to the largest cluster. (3) Scene coefficient fusion calculation: The number of clusters of all large-scale blocks is averaged to obtain the average number of clusters C. The specific calculation formula of the final scene coefficient S is as follows: ； Where Cmax is the preset maximum cluster size threshold, Rmax is 100%, S ranges from [0,1], and R represents the average percentage. For open scenes, S≥0.7, and for dense scenes, S≤0.

4. and These represent the proportion coefficients of the average percentage and the average number of clusters, respectively.

3. The high-definition video transmission method based on diversified data compression according to claim 1, characterized in that, Step S2 includes: An adaptive monitoring interval is established based on the scene coefficient, which has a value range of [0,1]. The video image processed in step S1 is divided into n interval levels. Different interval levels correspond to different monitoring interval lengths. The number of all monitoring interval lengths in the video image is counted, and the mode of the monitoring interval length is used as the final adaptive monitoring interval length. Within the adaptive monitoring interval, an inter-frame estimation algorithm is designed to calculate the sum of the absolute values ​​of the pixel grayscale differences between the frame to be detected and the reference frame, where the frame to be detected is the frame following the next frame between two adjacent frames, and the reference frame is the frame preceding the next frame between two adjacent frames. The similarity between the two frames is quantified, and the specific formula is as follows: + ； Where D represents the sum of the absolute values ​​of the pixel grayscale differences. Let be the pixel grayscale value at position (i,j) of the frame to be detected. Let be the pixel grayscale value at position (i,j) of the reference frame, and M and N be the height and width of the image, respectively. This represents the deviation of the cluster center coordinates between the frame to be detected and the reference frame, where || represents the absolute value calculation. The sum of the absolute values ​​of the pixel grayscale differences is normalized, and the inter-frame estimation coefficients of adjacent video images are calculated. The range of the inter-frame estimation coefficients is [0,1]. The closer the coefficient is to 0, the higher the similarity between the frame to be detected and the reference frame, and the greater the redundancy. The closer the coefficient is to 1, the lower the similarity between the frame to be detected and the reference frame, and the smaller the redundancy. The dynamic estimation threshold is initialized based on the scene complexity of the video image. The calculated inter-frame estimation coefficients are compared with the dynamic estimation threshold. If the inter-frame estimation coefficients are less than or equal to the threshold, the frame to be detected is determined to be a redundant frame and deleted. If the inter-frame estimation coefficients are greater than the threshold, the frame to be detected is determined to be a valid frame and retained. The process involves comparing all video images within the adaptive monitoring interval, identifying and deleting redundant frames, and obtaining an inter-frame video image set.

4. The high-definition video transmission method based on diversified data compression according to claim 3, characterized in that, The deviation value of the cluster center coordinates between the frame to be detected and the reference frame is calculated based on the coordinates of the center points between the clusters. The specific steps are as follows: (1) Use the DBSCAN algorithm to calculate the number of clusters of the frame to be detected and the number of pixels contained in each cluster. Use the DBSCAN algorithm to calculate the number of clusters of the reference frame and the number of pixels contained in each cluster. (2) Set the cluster set of the frame to be detected and the cluster set of the reference frame. Use the inter-frame matching algorithm to calculate the minimum value of the cluster center point coordinates in the cluster set of the frame to be detected and the cluster set of the reference frame. The minimum value is the deviation value of the cluster center point coordinates between the frame to be detected and the reference frame.

5. The high-definition video transmission method based on diversified data compression according to claim 1, characterized in that, Step S3 includes: (1) Perform grayscale gradient calculation on single-frame images in the inter-frame video image set, use the Sobel operator to extract the horizontal and vertical gradients of the image, calculate the texture direction and texture intensity of the pixel through the gradient value, and divide the image into three types of regions based on the texture direction and texture intensity: smooth region, edge region, and complex texture region. (2) Calculate the classification coefficient and clustering coefficient. The classification coefficient is the proportion of each region in the image. The K-Medoids clustering algorithm is used to calculate the clustering coefficient, which is the number of pixels contained in each cluster. Initialize the large clustering threshold. For clusters with a clustering coefficient greater than the large clustering threshold, they are marked as large clusters. (3) Use the classification coefficient to compress and reorganize the large clusters to obtain the intra-frame compressed set.

6. A high-definition video transmission system based on diversified data compression, characterized in that, include: The preprocessing module standardizes the video images, removes noise from the video images using Gaussian filtering, performs grayscale processing, designs a large-scale feature extraction algorithm based on the grayscale of the video images, and calculates scene coefficients. The inter-frame compression module establishes an adaptive monitoring interval based on scene coefficients, designs an inter-frame estimation algorithm within the adaptive monitoring interval, calculates the inter-frame estimation coefficients of adjacent video images, initializes a dynamic estimation threshold, and removes redundant video images based on the inter-frame estimation coefficients and the dynamic estimation threshold to obtain an inter-frame video image set. The intra-frame compression module designs an intra-frame prediction algorithm for the inter-frame video image set. It uses the texture features of the video images to calculate the classification coefficient and the clustering coefficient of each classification coefficient, and compresses and reassembles the large clusters in the video images to obtain the intra-frame compressed image set. The compensation compression and output module randomly selects adjacent video images from the intra-frame compressed image set, designs a comprehensive compression evaluation algorithm, calculates the intra-frame compression ratio, initializes a compression ratio threshold, performs compensation compression if the intra-frame compression ratio is less than the compression ratio threshold, and does not perform compensation compression if the intra-frame compression ratio is greater than or equal to the compression ratio threshold, thus obtaining a compressed video image set; extracts a subset of similar video images from the compressed video image set, reassembles the similar video image subsets to obtain a video output subset, and transmits the entire video output subset; A high-definition video transmission method based on diversified data compression as described in any one of claims 1-5 is implemented.

Images