Video coding method for quickly exporting video data

Abstract

The invention relates to the technical field of video coding, in particular to a video coding method for quickly exporting video data, and the method comprises the steps: collecting the video data, and carrying out the preprocessing; calculating a motion vector and a motion fluctuation coefficient of each pixel point according to the motion change of the pixel points in the front and back frame images; then calculating a texture oscillation weight by using the motion condition and the texture features of the pixel points; gaussian background modeling is carried out on video data, and video coding is carried out through the ROI coding technology. The invention aims to realize high-precision extraction of a background region in a dynamic environment, enhance the ROI extraction quality in the dynamic environment and ensure the ROI coding quality of video data in a complex motion scene.

Description

Technical Field

[0001] The present application relates to the field of video encoding technology, and in particular to a video encoding method for fast exporting of video data. Background Art

[0002] Against the backdrop of the rapid development of digital media technologies, the acquisition, processing, and transmission of video data have become core components in various application scenarios. With the widespread adoption of HD, UHD, and even 8K video content, the data volume of video files has grown exponentially, placing higher demands on storage space and transmission bandwidth. In this context, efficient video coding technology has become a key enabler for the rapid export, compression, and distribution of video data. Current mainstream video coding standards, such as H.264 and H.265, effectively eliminate redundant video information through various methods such as intra-frame prediction and inter-frame prediction, significantly improving compression efficiency. Region of Interest (ROI) coding is a commonly used and effective mainstream technology that can significantly improve video coding quality.

[0003] ROI encoding technology differentially encodes key areas in a video, reducing the overall bitrate while maintaining subjective visual quality. However, existing ROI extraction methods primarily rely on low-level visual features, such as motion vectors and texture contrast, treating the moving target area as the region of interest, which is easily affected by external factors. In the case of camera motion or dynamic scene changes, the relative motion of the foreground and background often results in both exhibiting significant motion characteristics and texture contrast changes. In such cases, using existing ROI extraction techniques will affect ROI extraction accuracy, which in turn affects video encoding quality. Summary of the Invention

[0004] In view of the above, it is necessary to provide a video encoding method for quickly exporting video data to solve the above problems.

[0005] One embodiment of the present application provides a video encoding method for quickly exporting video data, the method comprising: Get all frame images in the video data; The optical flow method is used to obtain the motion vector of each pixel between two adjacent frames. For each frame and all previous frames, the horizontal variation characteristics of the motion vector of each pixel between two adjacent frames are analyzed to obtain the overall motion vector of each pixel in each frame. For each frame and the previous frame, the module length of the motion vector of each pixel is compared with the module length of the overall motion vector. The distribution characteristics of the module length of the overall motion vector obtained for each frame and all previous frames are combined to obtain the motion fluctuation coefficient of each pixel in each frame. Obtain the LBP feature map of each frame of image, analyze the change characteristics of the overall motion vector of each pixel point in each frame of image and the previous frames of images, obtain the pixel points pointed to by each pixel point in each frame of image in the previous frames of images, compare the LBP codes of each pixel point with its pointed pixel points, and combine the motion fluctuation coefficient to determine the texture oscillation weight of each pixel point in each frame of image; Based on the distribution of the texture oscillation weights of each pixel point in each frame of image, confirm the learning rate of each pixel point in each frame of image, use Gaussian background modeling to obtain the foreground mask map of each frame of image, and combine the ROI coding technology to encode the video data.

[0006] Among them, the obtaining of the overall motion vector of each pixel point in each frame of image is specifically as follows: Obtain the horizontal direction vector of the motion vector between adjacent two frames of images for each pixel; for each frame of image and all the previous frames of images, obtain the cumulative vector of the horizontal direction vectors between each pixel point in all adjacent frames of images as the overall motion vector of each pixel point in each frame of image.

[0007] Among them, the obtaining of the motion fluctuation coefficient of each pixel point in each frame of image is specifically as follows: Obtain the modulus length of the motion vector between the previous frame of each pixel point in each frame of image and each frame of image, and the modulus length of the motion vector in the horizontal direction, which are respectively denoted as the first modulus length and the second modulus length; calculate the difference between the first modulus length and the second modulus length; obtain the dispersion degree of the modulus lengths of the overall motion vectors of each pixel point in each frame of image and all the previous frames of images; use the result of positive fusion of the difference and the dispersion degree as the motion fluctuation coefficient of each pixel point in each frame of image.

[0008] Among them, the dispersion degree of the modulus length is calculated by variance.

[0009] Among them, the obtaining of the pixel points pointed to by each pixel point in each frame of image in the previous frames of images is specifically as follows: Calculate the difference vector between the overall motion vectors of each pixel point in each frame of image and its previous frames of images, and in the previous frames of each frame of image, starting from each pixel point, obtain the pointed pixel points according to the direction and modulus length of the corresponding difference vector.

[0010] Among them, the determination of the texture oscillation weight of each pixel point in each frame of image is specifically as follows: Obtain the decimal representation of the LBP code of each pixel point in each frame of image, denoted as the first representation value; Denote the decimal representation of the LBP code of the pixel points pointed to by each pixel point in each frame of image in the previous frames of images as the second representation value; Before obtaining each frame of image, obtain the ratio of the second representation value pointing to the pixel in each previous frame of image to the first representation value of each pixel in each frame of image, calculate the absolute value of the difference obtained by subtracting 1 from the ratio, and use the result of positively fusing the mean value of all non-zero absolute values of the differences obtained and the motion fluctuation coefficient of each pixel in each frame of image as the texture oscillation weight of each pixel in each frame of image.

[0011] Among them, the specific formula of the texture oscillation weight is: Denote the texture oscillation weight of pixel p in the i-th frame of image as , and its formula form is: ; In the formula, represents the motion fluctuation coefficient of pixel p in the i-th frame of image; represents the decimal representation of the LBP encoding pointing to the pixel in the n-th frame of image before the i-th frame of image; represents the decimal representation of the LBP encoding value of pixel p in the i-th frame of image; i represents the frame number corresponding to each frame of image; u represents the number of non-zero ones.

[0012] Among them, when there is no pixel point pointed to in a previous frame of image for each pixel point in each frame of image, use the first representation value as the corresponding second representation value.

[0013] Among them, the method for determining the learning rate of each pixel in each frame of image is specifically: normalize the texture oscillation weight of each pixel in each frame of image, and use the difference between 1 and the obtained normalization result as the learning rate of the corresponding pixel.

[0014] Among them, the specific process of using Gaussian background modeling to obtain the foreground mask image of each frame of image and encoding the video data in combination with ROI encoding technology is as follows: Use each frame of image, the learning rate of each pixel in the image, and the background model obtained from the previous frame of image as inputs, and use Gaussian background modeling to output the foreground mask image of the current frame and the updated background model; Use the foreground area as the ROI, use each frame of image of the video data and the foreground mask image of each frame of image as inputs, and use ROI encoding technology to output the encoded bitstream.

[0015] This application has at least the following beneficial effects: First, this application calculates the motion vector and motion fluctuation coefficient of each pixel point through the motion characteristics of pixel points in adjacent front and rear frame images. The motion vector measures the cumulative motion of pixel points, which helps improve the accuracy of subsequent pixel texture change measurement by combining motion characteristics. The motion fluctuation coefficient enhances the accurate quantification of the motion characteristics of foreground and background pixel points. Then, using the motion situation and the texture characteristics of pixel points, the texture oscillation weight is calculated. This index provides motion and texture feature information for subsequent background extraction, which helps enhance the accuracy of background extraction. In this way, for the situation where there is relative motion in the background, the motion characteristics of pixel points and the corresponding relationship between motion and pixel texture change are introduced into the background extraction of video data, thereby achieving high-precision extraction of the background area in a dynamic environment, enhancing the extraction quality of the ROI in a dynamic environment, and ensuring the ROI coding quality of video data in a complex motion scene. Description of the Drawings

[0016] Figure 1 It is a flowchart of the video coding method for fast export of video data provided by this application; Figure 2 It is a schematic diagram of obtaining the learning rate of each pixel point of each frame image provided by this application. Detailed Implementation Manner

[0017] In the description of the embodiments of this application, words such as "exemplary", "or", "for example" are used to indicate examples, illustrations or explanations. Any embodiment or design solution described as "exemplary" or "for example" in the embodiments of this application should not be construed as being more preferred or having more advantages than other embodiments or design solutions. Rather, the use of words such as "exemplary", "or", "for example" is intended to present related concepts in a specific manner.

[0018] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the technical field to which this application belongs. The terms used in the description of this application are only for the purpose of describing specific embodiments and are not intended to limit this application.

[0019] In addition, it should be noted that the terms "first" and "second" in this application and its drawings are used to distinguish similar objects, rather than to describe a specific order or sequence. For the method disclosed in the embodiments of this application or the method shown in the flowchart, including one or more steps for implementing the method, without departing from the scope of protection of this application, the execution order of multiple steps can be interchanged with each other, and some steps can also be deleted.

[0020] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those of ordinary skill in the technical field to which this application pertains.

[0021] This application proposes a video encoding method for rapid export of video data, which is applied to the field of video encoding technology. Referring to the attached Figure 1 , the method includes the following steps: S1: Obtain all frame images in the video data.

[0022] Collect video data using a video camera with a frame rate of 60fps. Take the collected video data as input and extract frames one by one to obtain the image of each frame. Since images are often processed in the form of YUV video images during encoding, the image of each frame is taken as input and converted to the YUV format respectively. Image format conversion is a well-known technology and will not be elaborated here.

[0023] S2: Use the optical flow method to obtain the motion vector of each pixel point between adjacent two-frame images. For each frame image and all previous frame images, analyze the change characteristics of the motion vector of each pixel point in the horizontal direction between adjacent two-frame images to obtain the overall motion vector of each pixel point in each frame image; for each frame image and the previous frame image, compare the modulus length of the motion vector of each pixel point with the modulus length of the overall motion vector, and combine the distribution characteristics of the modulus length of the overall motion vector obtained from each frame image and all previous frame images to obtain the motion fluctuation coefficient of each pixel point in each frame image.

[0024] Since there is relative motion between the foreground and the background, both the foreground and the background have motion characteristics, but there are certain differences in their motion characteristics. First, the background can be considered fixed. Only when the foreground has active motion, the background can only move passively. Therefore, when the foreground has active motion, it can move in any direction, while the passive motion of the background only has horizontal motion, that is, obvious motion changes can only occur when the motion direction of the foreground is not perpendicular to the horizontal direction. Second, the background motion will cause changes in the pixels in the image, and motion will also generate motion vectors, which describe the motion direction and motion distance; for any two-frame images in the video data, the closer the corresponding motion vectors are, the higher the similarity between the pixel texture features in the two-frame images. Finally, since in most cases the movement of the camera follows the foreground objects, in some cases, the objects in the foreground have obvious motion, but the background motion is relatively small or does not move, that is, the fluctuation of the background motion is significantly less than that of the foreground motion.

[0025] Each frame of the video data is transformed into a grayscale image respectively, and then any two adjacent grayscale images are used as inputs. The optical flow method is adopted to output the corresponding optical flow field, and the motion vector of each pixel in the optical flow field between two adjacent frames is obtained and decomposed into a horizontal direction vector and a vertical direction vector. Among them, the optical flow method and vector decomposition are well-known technologies and will not be elaborated here. According to the above method, the optical flow fields between all adjacent frames of images are obtained respectively.

[0026] Based on the above analysis, the overall motion vector and motion fluctuation coefficient of each pixel point are calculated to measure the overall motion of each pixel point in the video data. Specifically: for each frame of image and all the previous frames of images, the cumulative vector of the horizontal direction vectors of each pixel point between all adjacent frames of images is obtained as the overall motion vector of each pixel point in each frame of image.

[0027] It should be noted that the overall motion vector of the pixel points between the current frame and the previous frame can be obtained only starting from the second frame of image.

[0028] It can be understood that since the motion of the pixel points in the video data is continuous, and for the background, the change in the image texture caused by the motion is mainly caused by the horizontal motion. That is, for the background pixel points, the motion distance in the horizontal direction is the main reason for the change in the pixel point texture. Therefore, the overall motion vector can measure the change of the pixel points. The closer the overall motion vectors of the pixel points between two frames of images are, the higher the similarity between the images.

[0029] Furthermore, the modulus length of the motion vector between the previous frame and the current frame of each pixel point in each frame of image and the modulus length of the motion vector in the horizontal direction are obtained, which are respectively denoted as the first modulus length and the second modulus length; the difference between the first modulus length and the second modulus length is calculated; the dispersion degree of the modulus length of the overall motion vector of each pixel point in each frame of image and all the previous frames of images is obtained; the result of the positive fusion of the difference and the dispersion degree is used as the motion fluctuation coefficient of each pixel point in each frame of image. In this embodiment, the difference between the modulus lengths corresponding to different vectors is calculated by the absolute value of the difference; the dispersion degree of the modulus lengths of the corresponding vectors of the pixel points in multiple images is calculated by variance; the difference of the pixel point p in the i-th frame of image is denoted as and the dispersion degree of the pixel point p in the i-th frame of image is denoted as The formula form of the motion fluctuation coefficient of the pixel point p in the i-th frame of image is: ; where e represents the natural constant.

[0030] It should be understood that for the background part in video data, only the horizontal motion can significantly affect the change of pixel points. Therefore, for the background motion, its motion vector is relatively close to the horizontal vector. Secondly, the amplitude and frequency of the background motion are relatively small, which is reflected as relatively small horizontal motion fluctuations, that is, relatively small motion vector fluctuations. Thus, the smaller the motion fluctuation coefficient, the more likely the pixel point belongs to the background area. On the contrary, due to the active motion characteristics of the foreground and the non-fixed motion direction, the corresponding motion fluctuation coefficient is relatively larger.

[0031] S3: Obtain the LBP feature map of each frame of image, analyze the change characteristics of the overall motion vector of each pixel point in each frame of image and the previous frames of images, obtain the pixel point pointed to by each pixel point in the previous frames of images for each frame of image, compare the LBP codes of each pixel point and its pointed pixel point, and combine the motion fluctuation coefficient to determine the texture oscillation weight of each pixel point in each frame of image.

[0032] When performing image background extraction, Gaussian background modeling is often used. It sets a baseline according to the pixel value of the pixel point. When the pixel values near the baseline randomly oscillate within a certain deviation, this part is the background area. However, when there is relative motion in the background, at this time, only the change of the pixel value is not enough to accurately depict the background characteristics again. Since there is motion in the background at this time, and the motion may also cause certain changes in the texture around the pixel points in the background, so at this time, it is necessary to consider the motion vector and texture change situation of each pixel point in the image, which helps to enhance the accuracy of subsequent background extraction.

[0033] At the same time, the motion vector can measure the motion direction and motion distance of pixel points in the image. In the image, since the foreground is actively moving and the motion mode is usually more complex, while the background is passively moving, the motion vector of the foreground is less regular compared with that of the background. That is to say, when any two frames of images have the same motion vector, it means that the backgrounds of the two frames of images are almost the same at this time, and the closer the motion vectors are, the higher the similarity of the backgrounds in the two frames of images. However, due to the usually more complex motion of the foreground object and the large changes in its position and shape in the image, even if the motion vectors of the two frames of images are close, it cannot ensure the similarity of the foreground.

[0034] Therefore, in order to enhance the accuracy of Gaussian background modeling for background extraction, the Gaussian background modeling process can be adjusted according to the relationship between texture change and motion situation, as well as the change characteristics of pixel points in the foreground and background. By introducing texture change and motion situation into the background extraction process, the accuracy of background extraction can be enhanced, which in turn helps to improve the quality of subsequent video coding.

[0035] Taking each frame image, sampling radius, and number of sampling points as inputs respectively, and using the local binary pattern, an LBP feature map is output. In this embodiment, the sampling radius and number of sampling points are set to 1 and 8 respectively. The local binary pattern is a well-known technology and will not be elaborated here.

[0036] Based on the above analysis, calculate the texture oscillation weight to measure the texture and motion of pixel points in each image for subsequent background extraction. Specifically: Obtain the decimal representation of the LBP encoding of each pixel point in each frame image, denoted as the first representation value; Calculate the difference vector between the overall motion vector of each pixel point in each frame image and the previous frame images, and in the previous frame images of each frame image, starting from each pixel point, according to the direction and magnitude of the corresponding difference vector, obtain the pointing pixel point of each pixel point in the previous frame images; Denote the decimal representation of the LBP encoding of the pointing pixel point of each pixel point in the previous frame images of each frame image as the second representation value; Obtain the ratio of the second representation value of the pointing pixel point in the previous frame images of each frame image to the first representation value of each pixel point in each frame image, calculate the absolute value of the difference between the obtained ratio minus 1, and take the result of the positive fusion of the mean of all non-zero absolute values of the differences and the motion fluctuation coefficient of each pixel point in each frame image as the texture oscillation weight of each pixel point in each frame image.

[0037] It should be noted that the pointing pixel point of pixel point p in the previous frame images of each frame image may not exist. When the pointing pixel point does not exist, take the first representation value as the corresponding second representation value.

[0038] In this embodiment, denote the texture oscillation weight of pixel point p in the i-th frame image as , and its formula form is: ; In the formula, represents the motion fluctuation coefficient of pixel point p in the i-th frame image; represents the decimal representation of the LBP encoding of the pointing pixel point in the n-th frame image before the i-th frame image; represents the decimal representation of the LBP encoding value of pixel point p in the i-th frame image; i represents the frame number corresponding to each frame image; u represents the number of non-zeros.

[0039] It can be understood that for background pixel points in the image, their motion changes show a relatively strict correspondence with the motion vector, and the motion fluctuation is relatively slight, so that the texture oscillation weight corresponding to the background pixel points is relatively small; on the contrary, for foreground pixel points, the regularity of the motion vector is poor, and the change of the foreground pixel points does not strictly correspond to the motion vector, so that the texture oscillation weight of the foreground pixel points is relatively large.

[0040] S4: Based on the distribution of the texture oscillation weights of each pixel in each frame of the image, determine the learning rate of each pixel in each frame of the image. Adopt Gaussian background modeling to obtain the foreground mask image of each frame of the image, and combine it with the ROI coding technology to encode the video data.

[0041] Calculate the texture oscillation weights of each pixel in each frame of the image respectively. The texture oscillation weights measure the texture changes and motion conditions of the pixels. When using Gaussian background modeling to construct the background model and extract the background subsequently, in traditional Gaussian background modeling, since the texture changes and motion conditions are not considered, the influence weights of any pixel are the same when constructing the background model. However, in the case of relative background motion, due to the obvious fluctuations of the background pixels, the probabilities of different pixels belonging to background pixels or foreground pixels are different. Traditional Gaussian background modeling will cause too large errors, thus affecting the accuracy of background extraction. Therefore, it is necessary to adjust the pixel weights according to the probabilities of different pixels belonging to background pixels or foreground pixels, and then control the learning rate of Gaussian background modeling through this weight, so as to achieve the control of the influence weights of different pixels.

[0042] In this embodiment, the formula for adaptive adjustment of the learning rate is: ; where is the adjusted learning rate of the pixel in the th frame of the image, is the texture oscillation weight of the pixel in the th frame of the image, and

[0043] is the normalization function. In different processing methods, it can adopt, including but not limited to, maximum value normalization, maximum-minimum value normalization. Figure 2 shown.

[0044] It can be understood that in Gaussian background modeling, when introducing a pixel into the background model, the learning rate needs to be used to control the influence weight of this pixel on the background model. When the learning rate is larger, the influence of this pixel on the background model is greater. And if the probability of a pixel being a background pixel is greater, the corresponding texture oscillation weight is smaller, so that the corresponding learning rate is larger, the influence on the background model is greater, and the corresponding pixel has a greater probability of being determined as a background pixel through the background model.

[0045] Obtain the learning rate of each pixel point in each frame of image respectively in the above manner. Then, take each frame of image, the learning rate of each pixel point in the image, and the background model obtained from the previous frame of image as inputs in sequence, and adopt Gaussian background modeling to output the foreground mask image of the current frame and the updated background model. Take the foreground region as the ROI, take each frame of image of the video data and the foreground mask image of each frame of image as inputs, and adopt ROI coding technology to output the encoded bitstream. Gaussian background modeling and ROI coding technology are well-known technologies and will not be elaborated here.

[0046] The flowcharts and block diagrams in the accompanying drawings illustrate the possible architectures, functions, and operations of systems, methods, and computer program products according to embodiments of the present application. In this regard, each block in the flowchart or block diagram may represent a module, a program segment, or a part of code, and the module, program segment, or part of code contains one or more executable instructions for implementing the specified logical function. In some alternative implementations, the functions marked in the block may occur in a different order than marked in the accompanying drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in the reverse order, which may depend on the functions involved. In the descriptions corresponding to the flowcharts and block diagrams in the accompanying drawings, the operations or steps corresponding to different blocks may also occur in a different order than disclosed in the descriptions. Sometimes, there is no specific order between different operations or steps. For example, two consecutive operations or steps may actually be executed substantially in parallel, and they may sometimes be executed in the reverse order, which may depend on the functions involved. Each block in the block diagram and / or flowchart, and the combination of blocks in the block diagram and / or flowchart, can be implemented by a dedicated hardware-based system for performing the specified functions or actions, or can be implemented by a combination of dedicated hardware and computer instructions.

[0047] The above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present application, and should all be included within the protection scope of the present application.

Claims

1. A video encoding method for fast export of video data, characterized in that, The method includes the following steps: Obtain all frame images in the video data; Use the optical flow method to obtain the motion vector of each pixel point between two adjacent frame images. For each frame image and all previous frame images, analyze the change characteristics of the motion vector of each pixel point in the horizontal direction between two adjacent frame images to obtain the overall motion vector of each pixel point in each frame image. For each frame image and the previous frame image, compare the modulus length of the motion vector of each pixel point with the modulus length of the overall motion vector, and combine the distribution characteristics of the modulus length of the overall motion vector obtained from each frame image and all previous frame images to obtain the motion fluctuation coefficient of each pixel point in each frame image; Obtain the LBP feature map of each frame image, analyze the change characteristics of the overall motion vector of each pixel point in each frame image and all previous frame images to obtain the pointing pixel point of each pixel point in each frame image in all previous frame images, compare the LBP codes of each pixel point and its pointing pixel point, and combine the motion fluctuation coefficient to determine the texture oscillation weight of each pixel point in each frame image; Based on the distribution of the texture oscillation weights of each pixel point in each frame image, confirm the learning rate of each pixel point in each frame image, use Gaussian background modeling to obtain the foreground mask map of each frame image, and combine the ROI coding technology to encode the video data.

2. The video encoding method for rapid export of video data according to claim 1, wherein The obtaining of the overall motion vector of each pixel point in each frame image is specifically as follows: Obtain the horizontal direction vector of the motion vector of each pixel between two adjacent frame images; for each frame image and all previous frame images, obtain the cumulative vector of the horizontal direction vectors of each pixel point between all adjacent frame images as the overall motion vector of each pixel point in each frame image.

3. The video encoding method for fast export of video data according to claim 1, characterized in that, The obtaining of the motion fluctuation coefficient of each pixel point in each frame image is specifically as follows: Obtain the modulus length of the motion vector between the previous frame image and each frame image of each pixel point in each frame image, and the modulus length of the motion vector in the horizontal direction, which are respectively denoted as the first modulus length and the second modulus length; calculate the difference between the first modulus length and the second modulus length; obtain the dispersion degree of the modulus length of the overall motion vector of each pixel point in each frame image and all previous frame images; use the result of positive fusion of the difference and the dispersion degree as the motion fluctuation coefficient of each pixel point in each frame image.

4. The video encoding method for fast export of video data according to claim 3, wherein The dispersion degree of the modulus length is calculated by variance.

5. The video encoding method for rapid export of video data according to claim 1, characterized in that, The obtaining of the pointing pixel point of each pixel point in each frame image in all previous frame images is specifically as follows: Calculate the difference vector between the overall motion vector of each pixel point in each frame image and all previous frame images, and in all previous frame images of each frame image, starting from each pixel point, obtain the pointing pixel point according to the direction and modulus length of the corresponding difference vector.

6. The video encoding method for fast export of video data according to claim 1, wherein, The determining of the texture oscillation weight of each pixel point in each frame image is specifically as follows: Obtain the decimal representation of the LBP code of each pixel point in each frame image, denoted as the first representation value; Denote the decimal representation of the LBP code of the pointing pixel point of each pixel point in each frame image in all previous frame images as the second representation value; Before obtaining each frame of image, obtain the ratio of the second representation value of the pixel point pointed to in each previous frame of image to the first representation value of each pixel point in each frame of image, calculate the absolute value of the difference obtained by subtracting 1 from the ratio, and use the result of positively fusing the mean value of all non-zero absolute values of the differences obtained with the motion fluctuation coefficient of each pixel point in each frame of image as the texture oscillation weight of each pixel point in each frame of image.

7. The video encoding method for fast export of video data according to claim 6, wherein, The specific formula for the texture oscillation weight is: Denote the texture oscillation weight of the pixel point p in the i-th frame image as , and its formula form is: ; In the formula, represents the motion fluctuation coefficient of the pixel point p in the i-th frame image; represents the decimal representation of the LBP encoding pointing to the pixel point in the n-th frame image before the i-th frame image; represents the decimal representation of the LBP encoding value of the pixel point p in the i-th frame image; i represents the frame number corresponding to each frame image; u represents the number of non-zero ones.

8. The video encoding method for fast export of video data according to claim 6, wherein When the pixel point pointed to in a previous frame of image for each pixel point in each frame of image does not exist, use the first representation value as the corresponding second representation value.

9. The video encoding method for fast export of video data according to claim 1, characterized in that, To confirm the learning rate of each pixel point in each frame of image, specifically: normalize the texture oscillation weight of each pixel point in each frame of image, and use the difference between 1 and the obtained normalization result as the learning rate of the corresponding pixel point.

10. The video encoding method for fast export of video data according to claim 1, characterized in that, The specific process of using Gaussian background modeling to obtain the foreground mask image of each frame of image and encoding the video data in combination with ROI encoding technology is as follows: Take each frame of image, the learning rate of each pixel point in the image, and the background model obtained from the previous frame of image as inputs, use Gaussian background modeling, and output the foreground mask image of the current frame and the updated background model; Take the foreground area as the ROI, take each frame of image of the video data and the foreground mask image of each frame of image as inputs, use ROI encoding technology, and output the encoded bitstream.

Images