Video black field detection method based on visual analysis

Abstract

The invention discloses a video black field detection method based on visual analysis, and the method comprises the following steps: A1, carrying out the brightness detection of a video image, and entering a step A2 if the brightness of the video image is continuously lower than a black field brightness threshold value in a judgment period, i.e., a brightness black field is established; a2, performing static frame detection on the video picture, if a static frame condition is met, entering a step A3, otherwise, skipping to the step A1; and A3, carrying out picture light abrupt change detection on the video picture, searching a point with the highest brightness abrupt change value, and if the brightness abrupt change value of the point is smaller than the brightness abrupt change threshold value, judging that the picture is a black field. According to the scheme, the brightness change caused by light reflection and a real black field picture can be effectively distinguished, and false alarms caused by light reflection are reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of video processing, and particularly to an enhanced algorithm for video black field detection based on visual analysis, which is used to improve the accuracy of video black field detection and reduce false alarm problems caused by factors such as reflection. Background Art

[0002] In the existing video black field detection technology, a detection algorithm based on the brightness of the picture is usually adopted. However, in the actual application process, since the source video signal is captured by a camera, the picture is easily affected by external light, especially the reflection phenomenon has a greater impact on the black field judgment algorithm. When strong reflection appears, the reflection light band is significantly white, and the brightness value is even higher than that of some normal pictures, resulting in very serious false alarms of the black field. The existing black field detection algorithms cannot effectively distinguish the brightness change caused by reflection from the real black field picture, thus unable to meet the requirements of high-precision black field detection. Summary of the Invention

[0003] The present invention mainly solves the technical problem of easy misjudgment existing in the prior art, and provides a video black field detection method based on visual analysis that can exclude reflection interference.

[0004] The present invention mainly solves the above technical problems through the following technical solutions: A video black field detection method based on visual analysis, including the following steps: A1. Perform brightness detection on the video picture. If the brightness of the video picture continuously remains lower than the black field brightness threshold within the determination period, then enter step A2; This is the basic step of black field detection, which is used to initially screen out possible black field pictures; A2. Perform still frame detection on the video picture. If the still frame condition is met, then enter step A3, otherwise jump back to step A1; Considering that the black field picture is usually accompanied by the establishment of the still frame condition, therefore, on the premise that the black field is established, verify whether the still frame condition is established to improve the accuracy of black field detection; A3. Perform picture light mutation detection on the video picture, find the point with the highest brightness mutation value. If the brightness mutation value of this point is less than the brightness mutation threshold, then determine that the picture is a black field. This detection step can effectively distinguish the brightness change caused by reflection from the real black field picture and reduce false alarms caused by reflection.

[0005] Only when the above three detection algorithms are all established, that is: the brightness detection determines as a black field, the still frame detection determines as a still frame, and the light mutation detection determines as no mutation, then finally determine that the current is a black field.

[0006] Preferably, before performing brightness detection on the video frame in step A1, the video frame is preprocessed, specifically including denoising, contrast enhancement, and region recognition. This solution is applicable to scenarios where the images of one or more display devices are captured by a high-definition monitoring camera and then it is determined whether there is a black field image. Therefore, in addition to traditional denoising and contrast enhancement, the images of each display device are divided through region recognition, and subsequent processing is also based on the images of individual display devices obtained through region recognition, so as to achieve simultaneous monitoring of multiple signals.

[0007] Preferably, in the step A1, the brightness B of the video frame t is calculated by the following formula: B t = mean(V(x,y)); V(x,y) is the brightness of the point with coordinates (x,y); B t is the average brightness of the entire frame; the determination period is 5 - 10 seconds.

[0008] Preferably, in the step A3, the brightness mutation value is determined by the following formula: f(x,y) = |V(x,y) - V(x + Δ,y + Δ)|; f(x,y) is the brightness mutation value of the point with coordinates (x,y); Δ is the offset, and the value range is [1,5]. If there is no corresponding point, the brightness mutation value is 0; the maximum value among all brightness mutation values is denoted as f max f max = max(f(x,y)); the brightness mutation threshold is 50. When f max is less than 50, it is determined that the mutation brightness of the frame is no mutation. In the traditional solution, the brightness mutation value is generally calculated through gradients. In this solution, the brightness mutation value is calculated by comparing with the points at a fixed offset position, which can significantly reduce the calculation amount, obtain a great advantage in speed, and can also be arranged on devices with relatively weak computing capabilities.

[0009] Preferably, in the step A2, the still frame determination method is: calculate the similarity S of the frames within the duration through the following formula frame : ; where, I j is the j-th frame image, SSD is the area of the different content blocks between the j-th frame image and the (j + 1)-th frame image / the total area of the frame, N is the number of frames analyzed, and the value of N is the black field determination period × the detection frequency. The detection frequency value range is [1,5]; if S frameIf it is less than the similarity threshold θ, it is considered a static frame. SSD compares the brightness of points at the same position in two frames of images. If the brightness difference is less than the brightness difference threshold, it is considered consistent, otherwise it is considered inconsistent. Then, it calculates the proportion of the number of inconsistent pixel points to the total number of pixel points. The value range of the brightness difference threshold is [15, 25]. The value range of the similarity threshold θ is [0.05, 0.15].

[0010] Preferably, in step A2, the static frame determination method can also be: S1. Determine whether there is a sample frame. If there is, extract a frame from the video as a comparison frame after delaying a sampling period t from the previous frame extraction time. If the sample frame does not exist, first extract the earliest picture in the video within the determination period as the sample frame, and then extract a picture as the comparison frame after delaying a sampling period t. If it is already outside the determination period after delaying a sampling period t, it is directly determined that it is not in the static frame state; S2. Divide the sample frame and the comparison frame pictures into several blocks of the same size; S3. Convert each pixel point in the entire pictures of the sample frame and the comparison frame into a brightness value in the range of [0, 255]. The brightness of the pixel point with coordinates (x, y) in the sample frame is denoted as V(x, y), and the brightness of the pixel point with coordinates (x, y) in the comparison frame is denoted as V'(x, y). For RGB images, the brightness value Gray is converted by Gray = 0.299×R + 0.587×G + 0.114×B; S4. Calculate the brightness difference of each pixel point between the sample frame and the comparison frame. The formula is as follows: V diff(x,y) = |V(x, y) - V'(x, y)|; V diff(x,y) is the brightness difference of the pixel points with coordinates (x, y) in both the sample frame and the comparison frame; S5. Count the pixel points with a brightness difference greater than the brightness difference threshold V Limit in each block. The total number is S. Divide S by the total number of pixels K in the block to obtain the difference degree data D between the blocks at the same position in the sample frame and the comparison frame. The value range of D is [0, 1]. 0 means the two blocks are exactly the same, and 1 means the two blocks are completely different; S6. Use the difference degree data corresponding to the block with the largest difference degree data as the difference degree value Diff max of the entire picture; When Diff max is greater than or equal to the static frame upper threshold Diff Limit1 , it is considered that the two frames of images are inconsistent. Take the comparison frame as the sample frame, set the static frame duration to 0, and jump to step S1 to restart the detection process; When Diffmax Greater than the still frame lower threshold Diff Limit2 and less than the still frame upper threshold Diff Limit1 When this is the case, it is considered that the image is in a blurred state, the sample frame remains unchanged, and the still frame duration increases by 0.5 time units T x , and then proceed to step S7; When Diff max is less than or equal to the still frame lower threshold Diff Limit2 it is considered that the two frames of images are the same, the sample frame remains unchanged, and the still frame duration increases by one time unit T x , and then proceed to step S7; S7. Determine whether the still frame duration exceeds the set still frame time limit. If it exceeds, it is determined that the still frame is established. If it does not exceed, jump to step S1 to restart the detection process.

[0011] Preferably, the sampling period t ranges from [200, 1000], with the unit of milliseconds; the length and width of each block range from [5, 30], with the unit of pixels; the still frame upper threshold Diff Limit1 ranges from [0.1, 0.3], the still frame lower threshold Diff Limit2 = Diff Limit1 - 0.05; the still frame time limit ranges from [5, 10] and does not exceed the determination period, with the unit of seconds; the brightness difference threshold V Limit ranges from [15, 25]. The smaller the value of the brightness difference threshold, the higher the detection sensitivity and the higher the false alarm rate.

[0012] Preferably, the time unit T x is determined by the following formula: T x = t * V Limit / 20. The value of the time unit T x is inversely correlated with the sensitivity of V Limit . When the sensitivity is higher (V Limit is smaller), the increase in each duration is smaller, so the time required to confirm the still frame is longer, which can effectively prevent the problem of increased false alarm rate caused by the increase in sensitivity.

[0013] Preferably, in the first five determinations, the difference degree data is calculated for all blocks, and the maximum value is selected as the difference degree value of the entire screen; in subsequent determinations, the average value of the difference degree data in the last five determinations of each block is used as the historical difference degree of this block, and then all blocks are sorted from largest to smallest according to the historical difference degree. The top 20% are used as the first-priority blocks, and the 8 blocks around each first-priority block (upper left, up, upper right, left, right, lower left, down, lower right, ignoring if the corresponding block does not exist) are used as the second-priority blocks. If a block meets both the first-priority block and the second-priority block, it is classified as a first-priority block; then, first calculate the difference degree data of all first-priority blocks, select the maximum value and compare it with the still-frame upper threshold. If the maximum difference degree data is greater than or equal to the still-frame upper threshold, it is determined that the two frames of images are inconsistent and corresponding operations are performed; if the maximum difference degree data is less than the still-frame upper threshold, calculate the difference degree data of all second-priority blocks, and select the maximum value and compare it with the still-frame upper threshold. If the maximum difference degree data is greater than or equal to the still-frame upper threshold, it is determined that the two frames of images are inconsistent and corresponding operations are performed; if the maximum difference degree data is less than the still-frame upper threshold, process according to step S5. Calculate the difference degree data in the manner of step S5. If the maximum difference degree data in the first-priority block or the second-priority block is greater than the still-frame upper threshold, the corresponding operation means that the sample frame and the comparison frame images are considered inconsistent, the comparison frame is used as the sample frame, the still-frame duration is set to 0, and the process jumps to step S1 to start the detection process again.

[0014] The higher the priority, the greater the possibility of picture change obtained from historical data, which can effectively reduce the time and computational complexity required for determination. Since the movement of objects in the vast majority of pictures is coherent, the surrounding blocks of each first-priority block are used as second-priority blocks, further reducing the occurrence probability of full-screen detection.

[0015] Preferably, if a block does not calculate the difference degree data in a certain round of determination (for example, if there is a difference degree data greater than the still-frame upper threshold in the first-priority block, the non-first-priority blocks will no longer calculate the difference degree data), the difference degree data of this block in this round is recorded as 0.

[0016] Each parameter involved in the solution is determined according to the usage scenario.

[0017] The substantial effects brought by the present invention are as follows: (1) Improving detection accuracy: By comprehensively considering black field brightness detection, still frame detection, and picture misalignment light mutation detection, the present invention can effectively improve the accuracy of black field detection, increasing the detection accuracy rate from 90% to 99.9%, and significantly reducing the occurrence of false alarms; (2) Enhancing robustness: The algorithm can adapt to video pictures under different lighting conditions. Even in complex environments such as strong reflection, it can accurately detect black field pictures, enhancing the robustness of black field detection; (3) Wide application range: The present invention is applicable to various video processing scenarios, such as television broadcasting, video surveillance, video editing, etc., and has broad application prospects and practical value. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 is a flowchart of a method for detecting black fields in videos according to the present invention; Figure 2 is a picture captured by a high-definition monitoring camera according to the present invention; Figure 3 is a picture obtained after area recognition and division according to the present invention; Figure 4 is another picture obtained after area recognition and division according to the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0019] The technical solutions of the present invention will be further specifically described below through embodiments in combination with the drawings.

[0020] Embodiment 1: A method for detecting black fields in videos based on visual analysis, as Figure 1 shown, includes the following steps: A1. Perform brightness detection on the video picture. If the brightness of the video picture continuously remains lower than the black field brightness threshold within the determination period, that is, the brightness black field is established, then proceed to step A2; This is the basic step of black field detection, used to initially screen out possible black field pictures; A2. Perform still frame detection on the video picture. If the still frame condition is met, then proceed to step A3, otherwise jump back to step A1; Considering that black field pictures usually accompany the establishment of the still frame condition, therefore, on the premise that the black field is established, verify whether the still frame condition is established to improve the accuracy of black field detection; A3. Perform picture light mutation detection on the video picture, and find the point with the highest brightness mutation value. If the brightness mutation value of this point is less than the brightness mutation threshold, then determine that the picture is a black field. This detection step can effectively distinguish the brightness change caused by reflection from the real black field picture, reducing false alarms caused by reflection.

[0021] Only when all the above three detection algorithms are established, that is: the brightness detection determines a black field, the still frame detection determines a still frame, and the light mutation detection determines no mutation, then finally determine that the current is a black field.

[0022] Before performing brightness detection on the video frame in step A1, preprocess the video frame, specifically including denoising, contrast enhancement, and region recognition. This solution is applicable to the scenario of capturing the frames of one or more display devices through a high-definition monitoring camera and then determining whether there is a black frame. Therefore, in addition to traditional denoising and contrast enhancement, region recognition is also used to divide the frames of each display device, so as to achieve simultaneous monitoring of multiple signals. Figure 2 It is an image obtained by grouping and monitoring the frames of multiple TV signals collected by a high-definition monitoring camera in this embodiment.

[0023] In step A1, the brightness Bt of the video frame is calculated by the following formula: Bt = mean(V(x, y)); V(x, y) is the brightness of the point with coordinates (x, y); the determination period is 5 - 10 seconds, and 7 seconds is taken in this embodiment.

[0024] In step A3, the brightness mutation value is determined by the following formula: f(x, y) = |V(x, y) - V(x + Δ, y + Δ)|; f(x, y) is the brightness mutation value of the point with coordinates (x, y); Δ is the offset, and the value range is [1, 5], and 3 is taken in this embodiment. If there is no corresponding point, the brightness mutation value is 0; the maximum value among all brightness mutation values is denoted as f max f max = max(f(x, y)); the brightness mutation threshold is 50. When f max is less than 50, it is determined that the picture has no brightness mutation. The traditional solution generally calculates the brightness mutation value through gradients. This solution calculates the brightness mutation value by comparing with the points at fixed offset positions, which can significantly reduce the calculation amount, obtain a great advantage in speed, and can also be arranged on devices with relatively weak computing capabilities.

[0025] In step A2, the still frame determination method is: calculate the picture similarity S within the duration through the following formula frame : ; where I j is the jth frame image, SSD is the area of the different content blocks between the jth frame image and the (j + 1)th frame image / the total area of the picture, N is the number of frames analyzed, and the value of N is the black frame determination period × the detection frequency. The detection frequency value range is [1, 5], and 3 is taken in this embodiment, that is, 3 pictures are taken per second for detection; if S frameIf it is less than the similarity threshold θ, it is considered a static frame. SSD compares the brightness of points at the same position in two frames of images. If the brightness difference is less than the brightness difference threshold, it is considered consistent; otherwise, it is considered inconsistent. Then, it calculates the ratio of the number of inconsistent pixel points to the total number of pixel points. The value range of the brightness difference threshold is [15, 25], and 20 is taken in this embodiment. The value range of the similarity threshold θ is [0.05, 0.15], and 0.1 is taken in this embodiment. Each parameter involved in the scheme can be determined according to the usage scenario and experience.

[0026] As Figure 3 shown is the picture of a block collected. When calculating the brightness result under strong reflection, the video picture brightness B t = 115.84, the black field brightness threshold is 120, and B t continuously remains less than the black field brightness threshold within the determination period; the static frame detection algorithm determines it as a static frame; the highest brightness mutation value f amx = 16, which is less than the mutation limit value of 50, and the picture light mutation determination condition is satisfied, so it is determined as a black field.

[0027] As Figure 4 shown is the picture of another block collected, which is a normal picture against a relatively black background. The video picture brightness B t = 76.71, the brightness limit value is 120, and the brightness black field is established; at the same time, the static frame algorithm determination is established; however, the highest brightness mutation value f amx = 252, and the mutation exceeds the limit value of 50, so the mutation condition is not satisfied, and it is determined as a non - black field. Figure 3 and Figure 4 has blurred the face.

[0028] Embodiment 2: A video black field detection method based on visual analysis. Except for the method of static frame determination, other processes are the same as those in Embodiment 1. The static frame determination method in this embodiment is as follows: S1. Determine whether there is a sample frame. If there is, extract a frame from the video as a comparison frame after delaying a sampling period t from the previous frame extraction time; if the sample frame does not exist, first extract the earliest picture from the video as the sample frame, and then extract a picture as the comparison frame after delaying a sampling period t; if it is already outside the determination period after delaying a sampling period t, directly determine that it is not in the static frame state; S2. Divide the sample frame and the comparison frame pictures into several blocks of the same size; S3. Convert each pixel in the entire images of the sample frame and the comparison frame into a brightness value in the range of [0, 255]. Denote the brightness of the pixel at coordinates (x, y) in the sample frame as V(x, y), and the brightness of the pixel at coordinates (x, y) in the comparison frame as V'(x, y); for RGB images, the conversion method for the brightness value Gray is Gray = 0.299×R + 0.587×G + 0.114×B; S4. Calculate the brightness difference of each pixel between the sample frame and the comparison frame. The formula is as follows: V diff(x,y) = |V(x, y) - V'(x, y)|; V diff(x,y) is the brightness difference of the pixel at coordinates (x, y) in both the sample frame and the comparison frame; S5. Count the number of pixels with a brightness difference greater than the brightness difference threshold V Limit in each block. The total number is S. Divide S by the total number of pixels K in the block to obtain the difference degree data D between the blocks at the same position in the sample frame and the comparison frame; the range of D is [0, 1], where 0 means the two blocks are exactly the same, and 1 means the two blocks are completely different; S6. Use the difference degree data corresponding to the block with the largest difference degree data as the difference degree value Diff max of the entire image; When Diff max is greater than or equal to the static frame upper threshold Diff Limit1 , it is considered that the two frames of images are inconsistent. Take the comparison frame as the sample frame, set the static frame duration to 0, and jump to step S1 to restart the detection process; When Diff max is greater than the static frame lower threshold Diff Limit2 and less than the static frame upper threshold Diff Limit1 , it is considered that the image is in a blurred state. Keep the sample frame unchanged, increase the static frame duration by 0.5 time units T x , and then enter step S7; When Diff max is less than or equal to the static frame lower threshold Diff Limit2 , it is considered that the two frames of images are consistent. Keep the sample frame unchanged, increase the static frame duration by one time unit T x , and then enter step S7; S7. Determine whether the static frame duration exceeds the set static frame time limit. If it exceeds, it is determined that the static frame is established; if it does not exceed, jump to step S1 to restart the detection process.

[0029] The value range of the sampling period t is [200, 1000], and in this embodiment, it is taken as 400, with the unit of millisecond; the value ranges of the length and width of each block are both [5, 30], and in this embodiment, it is taken as 15, with the unit of pixel; the upper threshold Diff Limit1 of the still frame has a value range of [0.1, 0.3], and in this embodiment, it is taken as 0.2, and the lower threshold Diff Limit2 of the still frame is Diff Limit1 - 0.05, that is, 0.15; the value range of the still frame time limit is [5, 10] and does not exceed the determination period, and in this embodiment, it is taken as 5, with the unit of second; the luminance difference threshold V Limit has a value range of [15, 25], and in this embodiment, it is taken as 20. The smaller the value of the luminance difference threshold, the higher the detection sensitivity and the higher the false alarm rate.

[0030] The time unit T x is determined by the following formula: T x = t * V Limit / 20. The time unit T x is inversely correlated with the V Limit sensitivity. When the sensitivity is higher (V Limit is smaller), the increase in the duration of each time is smaller, so the time required for still frame confirmation is longer, which can well prevent the problem of the increase in the false alarm rate caused by the increase in sensitivity.

[0031] In the first five determinations at the beginning, the difference degree data is calculated for all blocks, and the maximum value is selected as the difference degree value of the entire screen. In subsequent determinations, the average value of the difference degree data in the last 5 determinations of each block is used as the historical difference degree of this block. Then, all blocks are sorted in descending order according to the historical difference degree. The top 20% are used as the first-priority blocks, and the 8 blocks around each first-priority block (upper left, up, upper right, left, right, lower left, down, lower right, ignoring if the corresponding block does not exist) are used as the second-priority blocks. If a certain block meets the criteria of both the first-priority block and the second-priority block, it is classified as the first-priority block. Then, first calculate the difference degree data of all first-priority blocks, select the maximum value and compare it with the still-frame upper threshold. If the maximum difference degree data is greater than or equal to the still-frame upper threshold, it is determined that the two frames of images are inconsistent and corresponding operations are performed. If the maximum difference degree data is less than the still-frame upper threshold, calculate the difference degree data of all second-priority blocks, and select the maximum value and compare it with the still-frame upper threshold. If the maximum difference degree data is greater than or equal to the still-frame upper threshold, it is determined that the two frames of images are inconsistent and corresponding operations are performed. If the maximum difference degree data is less than the still-frame upper threshold, the process proceeds according to step S5. The difference degree data is calculated in the manner described in step S5. If the maximum difference degree data in the first-priority block or the second-priority block is greater than the still-frame upper threshold, the corresponding operation means that it is considered that the sample frame and the comparison frame images are inconsistent, the comparison frame is used as the sample frame, the still-frame duration is set to 0, and the process jumps back to step S1 to start the detection process again.

[0032] The higher the priority, the greater the possibility of a screen change obtained from historical data, thereby effectively reducing the time and computational complexity required for determination. Since the movement of objects in the vast majority of screens is coherent, the surrounding blocks of each first-priority block are used as the second-priority blocks, further reducing the occurrence probability of full-screen detection.

[0033] If the difference degree data of a certain block is not calculated in a certain round of determination (for example, if there is a difference degree data greater than the still-frame upper threshold in the first-priority block, the difference degree data of non-first-priority blocks will no longer be calculated), the difference degree data of this block in this round is recorded as 0.

[0034] All parameters involved in the solution are determined according to the usage scenario.

[0035] The specific embodiments described herein are merely illustrative of the spirit of the present invention. Those skilled in the art of the present invention can make various modifications or supplements to the described specific embodiments or use similar ways to replace them, but will not deviate from the spirit of the present invention or exceed the scope defined by the appended claims.

[0036] Although terms such as brightness detection, still frame detection, and light mutation detection are used more frequently in this article, the possibility of using other terms is not excluded. The use of these terms is only for the convenience of describing and explaining the essence of the present invention; interpreting them as any additional limitation is contrary to the spirit of the present invention.

Claims

1. A method for detecting video black fields based on visual analysis, characterized in that, It includes the following steps: A1. Perform brightness detection on the video frame. If the brightness of the video frame continuously drops below the black field brightness threshold within the determination period, go to step A2; A2. Perform still frame detection on the video frame. If the still frame condition is met, go to step A3; otherwise, jump back to step A1; A3. Perform sudden change detection on the light of the video frame to find the point with the highest brightness mutation value. If the brightness mutation value of this point is less than the brightness mutation threshold, determine that the frame is a black field.

2. The method for detecting video black field based on visual analysis according to claim 1, wherein Before performing brightness detection on the video frame in step A1, preprocess the video frame, specifically including denoising, contrast enhancement, and region recognition.

3. A method for detecting video black field based on visual analysis according to claim 1 or 2, characterized in that, In the step A1, the brightness B of the video picture t is calculated by the following formula: B t = mean(V(x,y)); V(x, y) is the brightness of the point with coordinates (x, y); the determination period is 5 - 10 seconds.

4. A method for detecting video black fields based on visual analysis according to claim 3, characterized in that, In step A3, the brightness mutation value is determined by the following formula: f(x, y)=|V(x, y)-V(x + Δ, y + Δ)|; f(x, y) is the brightness mutation value of the point with coordinates (x, y); Δ is the offset, and its value range is [1, 5]. If there is no corresponding point, the brightness mutation value is 0; the brightness mutation threshold is 50.

5. A method for detecting video black fields based on visual analysis according to claim 1, characterized in that, In the said step A2, the still frame determination method is: calculate the picture similarity S within the duration through the following formula frame :[[]]END]] ； Among them, I j is the j-th frame image, SSD is the area of the different content blocks between the j-th frame image and the (j + 1)-th frame image / the total area of the screen, N is the number of frames analyzed, the value of N is the black field determination period × the detection frequency, and the value range of the detection frequency is [1, 5]; if S frame is less than the similarity threshold θ, it is considered a still frame.

6. The method for detecting a video black field based on visual analysis according to claim 1, characterized in that, In step A2, the still frame determination method is as follows: S1. Determine whether there is a sample frame. If there is, extract a frame from the video as a comparison frame after delaying a sampling period t from the previous frame extraction moment; if there is no sample frame, first extract the earliest frame from the video as the sample frame, and then extract a frame as the comparison frame after delaying a sampling period t. If it is already outside the determination period after delaying a sampling period t, directly determine that it is not in the still frame state; S2. Divide the sample frame and the comparison frame into several blocks of the same size; S3. Convert each pixel point in the entire sample frame and the comparison frame into a brightness value in the range of [0, 255]. The brightness of the pixel point with coordinates (x, y) in the sample frame is denoted as V(x, y), and the brightness of the pixel point with coordinates (x, y) in the comparison frame is denoted as V'(x, y); S4. Calculate the brightness difference of each pixel point between the sample frame and the comparison frame. The formula is as follows: V diff(x,y) = |V(x, y) - V’(x, y)|; V diff(x,y) is the luminance difference of the pixel point where the coordinates of the sample frame and the comparison frame are both (x, y); S5. Count the pixel points in each block whose brightness difference is greater than the brightness difference threshold V Limit The total number is S. Divide S by the total number of pixels K in the block to obtain the difference data D between the same-position blocks of the sample frame and the comparison frame. The range of D is [0, 1], where 0 means the two blocks are exactly the same and 1 means the two blocks are completely different. S6. Use the difference data corresponding to the block with the largest difference data as the difference value Diff of the entire screen max ; When Diff max is greater than or equal to the still frame upper threshold Diff Limit1 it is considered that the two frames of images are inconsistent. The comparison frame is taken as the sample frame, the still frame duration is set to 0, and the process jumps to step S1 to start the detection process again; When Diff max is greater than the lower threshold Diff in the still frame Limit2 and less than the upper threshold Diff in the still frame Limit1 it is considered that the image is in a blurred state, the sample frame remains unchanged, and the duration of the still frame is increased by 0.5 time units T x , and then proceed to step S7; When Diff max is less than or equal to the still-frame threshold Diff Limit2 it is considered that the two frames of images are identical, the sample frame remains unchanged, and the duration of the still frame increases by a time unit T x , and then step S7 is entered; S7. Determine whether the still frame duration exceeds the set still frame time limit. If it exceeds, determine that the still frame is established; if it does not exceed, jump to step S1 to restart the detection process.

7. The method for detecting black field in video based on visual analysis according to claim 6, wherein, The value range of the sampling period t is [200, 1000], with the unit of millisecond; the value ranges of the length and width of each block are both [5, 30], with the unit of pixel; the upper threshold Diff Limit1 of the still frame has a value range of [0.1, 0.3], and the lower threshold Diff Limit2 of the still frame = Diff Limit1 - 0.05; the value range of the still frame time limit is [5, 10] and does not exceed the determination period, with the unit of second; the brightness difference threshold V Limit has a value range of [15, 25].

8. A method for detecting video black fields based on visual analysis according to claim 7, characterized in that, The time unit T x is determined by the following formula: T x = t * V Limit / 20。 9. A method for detecting video black field based on visual analysis according to claim 6, characterized in that, In the first five determinations at the beginning, the difference degree data is calculated for all blocks, and the maximum value is selected as the difference degree value of the entire screen; in subsequent determinations, the average value of the difference degree data at the last 5 determinations of each block is used as the historical difference degree of this block, and then all blocks are sorted in descending order according to the historical difference degree. The top 20% are used as the first-priority blocks, and the 8 blocks around each first-priority block are used as the second-priority blocks. If a certain block meets both the first-priority block and the second-priority block at the same time, it is classified as a first-priority block; then, the difference degree data of all first-priority blocks is calculated first, and the maximum value is selected and compared with the still-frame upper threshold. If the maximum difference degree data is greater than or equal to the still-frame upper threshold, it is determined that the two frames of images are inconsistent and corresponding operations are performed; if the maximum difference degree data is less than the still-frame upper threshold, the difference degree data of all second-priority blocks is calculated, and the maximum value is selected and compared with the still-frame upper threshold. If the maximum difference degree data is greater than or equal to the still-frame upper threshold, it is determined that the two frames of images are inconsistent and corresponding operations are performed; if the maximum difference degree data is less than the still-frame upper threshold, the process proceeds according to step S5.

10. A method for detecting video black fields based on visual analysis according to claim 9, characterized in that, If the difference degree data of a certain block is not calculated during a certain round of determination, the difference degree data of this block in this round is recorded as 0.

Images