A film archive patch damage detection method based on VSF visual saliency map

Abstract

The purpose of the present invention is to provide a film archive plaque damage detection method based on VSF visual saliency map, which uses a method combining frame difference method with visual saliency map, and only needs to calculate the saliency map of the suspicious area after the frame difference method is processed, and then double threshold determination is performed on the suspicious area to quickly detect whether there is plaque damage in the frame image. This invention can improve the algorithm speed, can better cope with large data volume detection tasks, and improve detection efficiency and accuracy.

Description

Technical Field

[0001] The present invention belongs to the technical fields of image recognition, archive big data processing, etc., and specifically relates to a film archive plaque damage detection method based on a VSF visual saliency map. Background Art

[0002] With the continuous development and application of digital technology, more and more archival materials are digitally preserved. However, there are still many old film archives that need to be digitized. These archives often have certain problems such as plaques and damage, which affect the effect and utilization value after digitization. Using computer vision and other technologies to realize the automatic detection of damaged patches in film archives can not only improve the detection efficiency, but also improve the detection accuracy and stability, and effectively ensure the recovery and utilization of film archives. However, the detection task of film archive image damage detection is usually large in data volume. After the image is digitized, the file is large and the number of frames is large. If the algorithm speed is too slow, it will consume huge computing power and affect the progress and efficiency of the entire detection task. Therefore, it is of great significance to combine visual saliency and frame difference method to simplify the detection model and reduce the calculation amount of the detection algorithm. Summary of the invention

[0003] The present invention proposes a film archive plaque damage detection method based on VSF visual saliency, which uses a method combining a frame difference method with a visual saliency map. It only needs to calculate the saliency map of the suspicious area after the frame difference method is processed, and then perform a double threshold judgment on the suspicious area to quickly detect whether there is plaque damage in the frame image. This invention can improve the algorithm speed, can better cope with large data volume detection tasks, and improve detection efficiency and accuracy.

[0004] Considering that plaque damage usually does not occur at the same position in two adjacent frames at the same time, in terms of the visual saliency of the frame to be detected, the visual saliency of the plaque area is obvious while the visual saliency of the adjacent frame in this area is not obvious. Therefore, the present invention adopts a method combining the frame difference method with the visual saliency map to make a preliminary prediction of the input film archive image, extract the suspicious area after the frame difference method is processed, and then combine the visual saliency map method to perform a double threshold judgment on the suspicious area to ensure the detection accuracy of the plaque damage frame, and then traverse all the pixels in the suspicious area to complete the plaque damage detection of the image frame. For the damage detection of the frame image, the present invention only needs to calculate the visual saliency map of the suspicious area obtained by the frame difference method, instead of calculating the visual saliency map of the entire frame image, so as to achieve good detection effect while reducing the calculation amount of detection and simplifying the detection model.

[0005] The technical solution specifically adopted by the present invention to solve the technical problem is:

[0006] A film archive patch damage detection method based on VSF visual saliency map comprises the following steps:

[0007] Step S1: Preliminary prediction of the plaque damage area is performed on the current frame image and its adjacent frames of the input film archive image using the frame difference method, and the suspicious area is extracted by selecting a threshold: if no suspicious area is detected, it is judged that this frame image is not damaged; if there is a suspicious area, the suspicious area is binarized, and then step S2 is executed;

[0008] Step S2: Calculate two visual saliency maps of the image obtained after binarization processing of the suspicious area of ​​the current frame;

[0009] Step S3: Perform threshold determination on each pixel in the suspicious area of ​​the calculated saliency map: if the threshold is not exceeded, the pixel is not a damaged pixel; if the threshold is exceeded, calculate the saliency of the pixel blocks of two adjacent frames, and mark the pixel blocks exceeding the threshold, and then execute step S4;

[0010] Step S4: Count the calibrated pixel blocks and perform threshold determination on the counting results again: if the number of counted pixel blocks is not less than the threshold, the pixel is not a damaged pixel; if the number of counted pixel blocks is less than the threshold, the pixel is determined to be a damaged pixel;

[0011] Step S5: Continue to execute step S3 until all suspicious areas of the saliency map are traversed;

[0012] Step S6: Integrate all damaged pixels after traversing the saliency map to obtain the patch damage detection result of the damaged frame of the film file.

[0013] Furthermore, in step S1, the frame difference method used is a method for detecting damage based on the frame difference between the plaque damage frame and the preceding and following frames, a total of three frames. The specific implementation steps are:

[0014] Step A1: First obtain the frame difference r between the tth frame and the t-1th frame b The frame difference between the tth frame and the t+1th frame is r f , the calculation formula is as follows:

[0015]

[0016] Among them, t, t+1, and t-1 represent the frame to be detected and its previous and next frames respectively, and r b Represents the difference between the frame to be detected and the previous frame, r f Indicates the difference between the frame to be detected and the next frame;

[0017] Step A2: Threshold processing is performed on the frame difference result, and the frame difference result r b and r f are greater than the threshold and the frame difference result r b and rf All pixels that are not 0 are set to 1 to obtain the frame difference result. The specific calculation formula is as follows:

[0018]

[0019] Wherein, sgn(·) represents the sign function, and d(x,y) is the suspicious area obtained after binarization of the frame difference result.

[0020] Further, in step S1, the adjacent frames of the shot boundary are obtained by using the shot boundary detection method, the shot boundary is divided into the head frame and the tail frame, the two consecutive frames after the head frame are selected as adjacent frames, and the two consecutive frames before the head frame are selected as adjacent frames; then, a video mutation shot boundary detection method based on color histogram is adopted to obtain the color histogram of the RGB components of the input video sequence, and calculate the frame difference of the color histogram of the front and back frames, and determine whether the shot is a boundary shot by counting the three-dimensional color value histogram difference z(k, k+1) of adjacent frames. If the histogram difference is greater than a certain value, it is a boundary shot. The calculation formula of the histogram difference z(k, k+1) is as follows:

[0021]

[0022] Among them, z(k,k+1) represents the frame difference, N is the total number of pixels in RGB color, and H K A three-dimensional color histogram of RGB.

[0023] Furthermore, in step S2, the visual saliency map is calculated by traversing the pixels of the binary image of the suspicious area, marking the pixel points with a pixel value of 1 and calculating the visual saliency sub-map of the pixel point, and finally superimposing the saliency sub-map of each marked pixel point to obtain the two visual saliency maps of the final image. The visual saliency map is a grayscale map with pixel values ​​between 0 and 1. The visual saliency sub-map calculation formula with (x, y) as the center pixel point is as follows:

[0024]

[0025] Among them, l∈{12,24,48,56,112} represents the area surrounded by a rectangular box with length and width of l. VSF off 、VSF on There are two saliency maps, center represents the central area of ​​the pixel, and surround represents the surrounding area of ​​the pixel, which are defined as:

[0026]

[0027] Where rectSum is the integral value of any rectangular area in the image;

[0028] Among them, two visual saliency maps VSFoff and VSF on The difference between the pixel value of the suspicious pixel and the pixel value of the surrounding rectangular area with a side length of l is taken as the average result. off and VSF on Used to detect black damaged plaques and white damaged plaques respectively.

[0029] Furthermore, in step S2, when calculating the saliency map, edge pixels are copied to fill the pixels, so as to prevent the rectangular frame l from exceeding the original resolution of the image when the center point is at the image boundary.

[0030] Furthermore, in step S3, threshold determination and pixel block counting are performed on the two saliency maps respectively, and the specific implementation steps are as follows:

[0031] Step B1: Set the threshold value th1;

[0032] Step B2: Determine pixel blocks in the saliency map Is there any pixel with a significance greater than the threshold th1? If the calculated result exceeds the threshold, the pixel coordinates are marked in the adjacent frame and the pixel block is calculated. and The saliency map of the image is 0, where the value of c is 0 and 1. A total of 18 pixel blocks need to be calculated for the saliency of adjacent frames.

[0033] Step B3: re-determine the pixel blocks exceeding the threshold th1 in the saliency map of the adjacent frame, and calibrate the pixel blocks exceeding the threshold; the definition of the pixel block is: is the rectangular area centered on pixel p(x,y) in the frame to be detected. and It is the pixel block at the same pixel position in adjacent frames, and the size is 5×5.

[0034] Further, in step S4, the calibrated pixel blocks in two adjacent frames are processed: for the 9 pixel blocks in one adjacent frame, set S t±1 , if there is a pixel block that has been calibrated, let S t±1 The statistical count is increased by one; the threshold th2 is set and the S of two adjacent frames is determined. t+1 , S t-1 Are they all less than the threshold th2? t+1 , S t-1 If both are smaller than the threshold th2, the pixel is considered to be a damaged pixel; otherwise, the pixel is not damaged.

[0035] Compared with the prior art, the present invention and its preferred solution only need to calculate the visual saliency map of the suspicious area obtained by the frame difference method for damage detection of frame images, rather than calculating the visual saliency map of the entire frame image. This achieves good detection effect while reducing the calculation amount of detection and simplifying the detection model. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] The present invention is further described in detail below with reference to the accompanying drawings and specific embodiments:

[0037] Figure 1 is a work flow chart of an embodiment of the present invention;

[0038] Figure 2 This is a suspicious area map and a detection result map obtained by using the frame difference method in an embodiment of the present invention. DETAILED DESCRIPTION

[0039] In order to make the features and advantages of this patent more obvious and easy to understand, the following embodiments are specifically described in detail as follows:

[0040] It should be noted that the following detailed descriptions are illustrative and are intended to provide further explanation of the present application. Unless otherwise specified, all technical and scientific terms used in this specification have the same meanings as those commonly understood by those skilled in the art to which the present application belongs.

[0041] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present application. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprise" and / or "include" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof.

[0042] This embodiment provides a film archive patch damage detection method based on VSF visual saliency, comprising the following steps:

[0043] Step S1: Use the frame difference method to preliminarily predict the plaque damage area of ​​the current frame image and its adjacent frames of the input film archive image, and select a suitable threshold to extract the suspicious area. If there is no suspicious area, the frame image is not damaged. If there is a suspicious area, the suspicious area is binarized, and then step S2 is executed;

[0044] In an implementation case of the present invention, in step S1, the frame difference method used is a method for detecting damage based on the frame difference between the plaque damage frame and the preceding and following frames, which is mainly based on the characteristics that the pixel movement of the preceding and following frames in the video sequence is small and the plaque is discontinuous in the space of the preceding and following frames, and can quickly detect the position of the plaque in the frame to be detected. The specific implementation steps are:

[0045] Step A1: first obtain the frame difference r between the tth frame and the t-1th frame b The frame difference between the tth frame and the t+1th frame is r f , the calculation formula is as follows:

[0046]

[0047] Among them, t, t+1, and t-1 represent the frame to be detected and its previous and next frames respectively, and r b Represents the difference between the frame to be detected and the previous frame, r f Indicates the difference between the frame to be detected and the next frame;

[0048] Step A2: threshold the frame difference result and convert the frame difference result r b and r f are greater than the threshold and the frame difference result r b and r f The pixels that are not 0 are set to 1, which is equivalent to binary processing, and the frame difference result can be obtained. The specific calculation formula is as follows:

[0049]

[0050] Wherein, sgn(·) represents the sign function, and d(x,y) is the suspicious area obtained after binarization of the frame difference result.

[0051] As a further preference, in step S1, the adjacent frames of the shot boundary are obtained by using a shot boundary detection method. A film archive image is generally composed of multiple shots. The shot boundary is divided into a head frame and a tail frame. The pixel difference between the tail frame of the previous shot and the head frame of the next shot is large. For the head frame, the two consecutive frames after the head frame will be selected as adjacent frames, and for the tail frame, the two consecutive frames before the head frame will be selected as adjacent frames. A method for detecting the boundary of a sudden change in video shot based on a color histogram is adopted. The color histogram of the RGB components of the input video sequence is calculated, and the frame difference of the color histogram of the previous and next frames is calculated. The difference z(k, k+1) of the three-dimensional color value histogram of adjacent frames is used to determine whether the shot is a boundary shot. If the histogram difference is greater than a certain value, it is a boundary shot, otherwise it is not. The calculation formula of the histogram difference z(k, k+1) is as follows:

[0052]

[0053] Among them, z(k,k+1) represents the frame difference, N is the total number of pixels in RGB color, and H K A three-dimensional color histogram of RGB.

[0054] Step S2, calculating two visual saliency maps of the image obtained after binarization processing of the suspicious area of ​​the current frame;

[0055] In an embodiment of the present invention, in step S2, the visual saliency map calculation method is to perform pixel traversal on the binary image of the suspicious area, mark the pixel points with a pixel value of 1 and calculate the visual saliency sub-map of this pixel point, and finally superimpose the saliency sub-map of each marked pixel point to obtain two visual saliency maps of the final image, and the visual saliency map is a grayscale map with pixel values ​​between 0 and 1. The visual saliency sub-map calculation formula with (x, y) as the center pixel point is as follows:

[0056]

[0057] Among them, l∈{12,24,48,56,112} represents the area surrounded by a rectangular box with length and width of l. VSF off 、VSF on There are two saliency maps, center represents the central area of ​​the pixel, and surround represents the surrounding area of ​​the pixel, which are defined as:

[0058]

[0059] Where rectSum is the integral value of any rectangular area in the image.

[0060] Among them, the two visual saliency maps VSF calculated off and VSF on The difference between the pixel value of the suspicious pixel and the pixel value of the surrounding rectangular area with a side length of l is taken as the average result. Therefore, VSF off and VSF on Used to detect black damaged plaques and white damaged plaques respectively.

[0061] Preferably, in step S2, when calculating the saliency map, the image border needs to be filled to prevent the rectangular frame l from partially exceeding the original resolution of the image when the center point is at the image border, thereby affecting the accuracy of the saliency map calculation. Therefore, the edge pixels are copied to fill the pixels.

[0062] Step S3, perform threshold determination on each pixel in the suspicious area of ​​the calculated saliency map. If it does not exceed the threshold, the pixel is not a damaged pixel. If it exceeds the threshold, calculate the saliency of the pixel blocks of two adjacent frames, and mark the pixel blocks that exceed the threshold, and then execute step S4;

[0063] In an implementation case of the present invention, in step S3, it is necessary to perform threshold determination and pixel block counting on the two saliency maps respectively, and the specific implementation steps are as follows:

[0064] Step B1, setting the threshold value th1;

[0065] Step B2: Determine pixel blocks in the salient map Is there any pixel with a significance greater than the threshold th1? If the calculated result exceeds the threshold, the pixel coordinates are marked in the adjacent frame and the pixel block is calculated. and The saliency map of the image is , where the value of c is 0 and 1, so the saliency of 18 pixel blocks in total needs to be calculated in adjacent frames;

[0066] Step B3: re-determine the pixel blocks in the adjacent frame saliency map that exceed the threshold th1, and calibrate the pixel blocks that exceed the threshold. The definition of the pixel blocks mentioned is: is the rectangular area centered on pixel p(x,y) in the frame to be detected. and It is the pixel block at the same pixel position in adjacent frames, and the size is 5×5.

[0067] Step S4, count the calibrated pixel blocks and perform threshold determination on the counting results again. If the number of counted pixel blocks is not less than the threshold, the pixel is not a damaged pixel. If the number of counted pixel blocks is less than the threshold, the pixel is determined to be a damaged pixel;

[0068] In an embodiment of the present invention, in step S4, the calibrated pixel blocks in two adjacent frames are processed. For the 9 pixel blocks in one adjacent frame, set S t±1 , if there is a pixel block that has been calibrated, let S t±1 The statistical count is increased by 1. Set the threshold th2 and determine the S of two adjacent frames. t+1 , S t-1 Are they all less than the threshold th2? t+1 , S t-1 If both are smaller than the threshold th2, the pixel is considered to be a damaged pixel; otherwise, the pixel is not damaged.

[0069] Step S5, continue to execute step S3 until all suspicious areas of the saliency map are traversed;

[0070] Step S6: After traversing the saliency map, all damaged pixels are integrated to obtain the patch damage detection result of the damaged frame of the film file.

[0071] According to the above scheme design, the following Figure 1 Further explanation of the specific implementation steps:

[0072] Step 1: Transfer the digitized film file image to the computer for processing;

[0073] Step 2: Use the frame difference method to extract the suspicious area of ​​the current frame. If the suspicious area cannot be extracted, the image of this frame is not damaged; if the suspicious area can be extracted, proceed to the next step;

[0074] Step 3: Binarize the suspicious area of ​​the current frame and calculate two visual saliency maps of the image. Then, perform threshold determination on each pixel in the suspicious area of ​​the calculated saliency map. If the determination result does not exceed the threshold value, the pixel is not damaged, and continue to step 5; if the evaluation result exceeds the threshold value, proceed to the next step.

[0075] Step 4: Calculate the saliency of pixel blocks in two adjacent frames, mark and count the blocks that exceed the threshold, and then perform threshold determination on the counting results. If the counting results of the two adjacent frames are not both less than the threshold, the pixel is not damaged, and continue to step 5; if the counting results of the two adjacent frames are both less than the threshold, mark it as a damaged pixel, record its (x, y) coordinates, and proceed to the next step;

[0076] Step 5: Determine whether the current suspicious pixel is the last suspicious pixel. If not, return to step 3 to determine the next pixel. If it is the last suspicious pixel, integrate the suspicious pixel detection results and their coordinates and output the damaged patch of the current frame.

[0077] Figure 2 The following are the suspicious area map and detection result map obtained by using the frame difference method, where Figure (a) is the original damaged frame image, and the plaque in the image is the damaged plaque to be detected. Figure (b) is the damaged plaque map finally detected by the method of the present invention. Figure (c) is the difference result extracted by the frame difference method, and Figure (d) is the result of selecting a suitable threshold to extract the image from the result obtained by the frame difference method, and then binarizing the image.

[0078] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0079] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0080] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0081] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0082] The above is only a preferred embodiment of the present invention, and does not limit the present invention in other forms. Any technician familiar with the profession may use the above disclosed technical content to change or modify it into an equivalent embodiment with equivalent changes. However, any simple modification, equivalent change and modification made to the above embodiment according to the technical essence of the present invention without departing from the technical solution of the present invention still belongs to the protection scope of the technical solution of the present invention.

[0083] This patent is not limited to the above-mentioned optimal implementation mode. Anyone can derive other forms of a film archive patch damage detection method based on VSF visual saliency map under the inspiration of this patent. All equal changes and modifications made according to the scope of the patent application of the present invention should be covered by this patent.

Claims

1. A film archive patch damage detection method based on VSF visual saliency map, characterized in that: The following steps are involved: Step S1: Preliminary prediction of the plaque damage area is performed on the current frame image and its adjacent frames of the input film archive image using the frame difference method, and the suspicious area is extracted by selecting a threshold: if no suspicious area is detected, it is judged that this frame image is not damaged; if there is a suspicious area, the suspicious area is binarized, and then step S2 is executed; Step S2: Calculate two visual saliency maps of the image obtained after binarization processing of the suspicious area of ​​the current frame; Step S3: Perform threshold determination on each pixel in the suspicious area of ​​the calculated saliency map: if the threshold is not exceeded, the pixel is not a damaged pixel; if the threshold is exceeded, calculate the saliency of the pixel blocks of two adjacent frames, and mark the pixel blocks exceeding the threshold, and then execute step S4; Step S4: Count the calibrated pixel blocks and perform threshold determination on the counting results again: if the number of counted pixel blocks is not less than the threshold, the pixel is not a damaged pixel; if the number of counted pixel blocks is less than the threshold, the pixel is determined to be a damaged pixel; Step S5: Continue to execute step S3 until all suspicious areas of the saliency map are traversed; Step S6: integrating all damaged pixels after traversing the saliency map to obtain the patch damage detection result of the damaged frame of the film file; In step S2, the visual saliency map is calculated by traversing the pixels of the binary image of the suspicious area, marking the pixel points with a pixel value of 1 and calculating the visual saliency sub-map of this pixel point, and finally superimposing the saliency sub-map of each marked pixel point to obtain two visual saliency maps of the final image. The visual saliency map is a grayscale map with pixel values ​​between 0 and 1. The calculation formula of the visual saliency sub-map with (x, y) as the center pixel point is as follows: Among them, l∈{12,24,48,56,112} represents the area surrounded by a rectangular box with length and width of l. VSF off 、VSF on There are two saliency maps, center represents the central area of ​​the pixel, and surround represents the surrounding area of ​​the pixel, which are defined as: Where rectSum is the integral value of any rectangular area in the image; Among them, two visual saliency maps VSF off and VSF on The difference between the pixel value of the suspicious pixel and the pixel value of the surrounding rectangular area with a side length of l is taken as the average result. off and VSF on Used to detect black damaged plaques and white damaged plaques respectively.

2. The method for detecting film archive patch damage based on VSF visual saliency map according to claim 1, characterized in that: In step S1, the frame difference method used is a method for detecting damage based on the frame difference between the plaque damage frame and the preceding and following frames, a total of three frames. The specific implementation steps are: Step A1: First obtain the frame difference r between the tth frame and the t-1th frame b The frame difference between the tth frame and the t+1th frame is r f , the calculation formula is as follows: Among them, t, t+1, and t-1 represent the frame to be detected and its previous and next frames respectively, and r b Represents the difference between the frame to be detected and the previous frame, r f Indicates the difference between the frame to be detected and the next frame; Step A2: Threshold processing is performed on the frame difference result, and the frame difference result r b and r f are greater than the threshold and the frame difference result r b and r f All pixels that are not 0 are set to 1 to obtain the frame difference result. The specific calculation formula is as follows: Wherein, sgn(·) represents the sign function, and d(x,y) is the suspicious area obtained after binarization of the frame difference result.

3. The method for detecting film archive patch damage based on VSF visual saliency map according to claim 1, characterized in that: In step S1, the adjacent frames of the shot boundary are obtained by using the shot boundary detection method, and the shot boundary is divided into the head frame and the tail frame. The two consecutive frames after the head frame are selected as adjacent frames, and the two consecutive frames before the head frame are selected as adjacent frames. Then, the video mutation shot boundary detection method based on the color histogram is used to calculate the color histogram of the RGB component of the input video sequence, and the frame difference of the color histogram of the previous and next frames is calculated. By counting the three-dimensional color value histogram difference z(k, k+1) of the adjacent frames, it is determined whether the shot is a boundary shot. If the histogram difference is greater than a certain value, it is a boundary shot. The calculation formula of the histogram difference z(k, k+1) is as follows: Among them, z(k,k+1) represents the frame difference, N is the total number of pixels in RGB color, and H k A three-dimensional color histogram of RGB.

4. The method for detecting film archive patch damage based on VSF visual saliency map according to claim 3, characterized in that: In step S2, when calculating the saliency map, edge pixels are copied to fill the pixels, so as to prevent the rectangular frame l from exceeding the original resolution of the image when the center point is at the image boundary.

5. The method for detecting film archive patch damage based on VSF visual saliency map according to claim 1, characterized in that: In step S3, threshold determination and pixel block counting are performed on the two saliency maps respectively. The specific implementation steps are as follows: Step B1: Set the threshold value th1; Step B2: Determine pixel blocks in the saliency map Is there any pixel with a significance greater than the threshold th1? If the calculated result exceeds the threshold, the pixel coordinates are marked in the adjacent frame and the pixel block is calculated. and The saliency map of the image is 0, where the value of c is 0 and 1. A total of 18 pixel blocks need to be calculated for the saliency of adjacent frames. Step B3: re-determine the pixel blocks exceeding the threshold th1 in the saliency map of the adjacent frame, and calibrate the pixel blocks exceeding the threshold; the definition of the pixel block is: is the rectangular area centered on pixel p(x,y) in the frame to be detected. and It is the pixel block at the same pixel position in adjacent frames, and the size is 5×5.

6. The method for detecting film archive patch damage based on VSF visual saliency map according to claim 5, characterized in that: In step S4, the calibrated pixel blocks in two adjacent frames are processed: for the 9 pixel blocks in one adjacent frame, set S t±1 , if there is a pixel block that has been calibrated, let S t±1 The statistical count is increased by one; the threshold th2 is set and the S of two adjacent frames is determined. t+1 , S t-1 Are they all less than the threshold th2? If S t+1 , S t-1 If both are smaller than the threshold th2, the pixel is considered to be a damaged pixel; otherwise, the pixel is not damaged.

Images