A method for detecting the quality of archive images

Abstract

The present invention belongs to the field of image processing technology, and specifically relates to a method for detecting the quality of an archival image, comprising: translating a connected domain of a target grayscale value in an archival image in different directions, obtaining a maximum matching rate between the target grayscale value and the remaining grayscale values during the translation, forming a translation feature quantity with the direction and step length corresponding to the maximum matching rate, using the grayscale value corresponding to the maximum matching rate as a matching grayscale value, clustering all grayscale values into multiple categories based on the translation feature quantity of each grayscale value, determining the grayscale concentration and matching grayscale concentration of each category, performing Gaussian fitting on the number of matching pixels corresponding to the maximum matching rate of each grayscale value in each category, determining the ghosting probability of each category based on the fitting error, grayscale concentration, matching grayscale concentration, and the maximum matching rate of the grayscale value, and screening the ghosting area. The present invention can accurately identify the ghosting area, thereby improving the accuracy and reliability of archival image quality detection.

Description

Technical Field

[0001] The present invention relates to the field of image processing technology, and more particularly to a method for detecting the quality of archive images. Background Art

[0002] Archive digitization is a crucial technological tool for preserving historical documents and improving information utilization efficiency. However, during the digital scanning process, factors such as mechanical vibration of the equipment, displacement of the scanning platform, tilted document placement, and improper operation often lead to ghosting in the scanned image. Ghosting primarily manifests as multiple offsets and superpositions of the outlines of the same text or pattern. This phenomenon significantly reduces image readability and negatively impacts subsequent optical character recognition, image analysis, and long-term preservation.

[0003] In the related art, for example, a Chinese patent document with authorization announcement number CN118865397B discloses a method and device for color text printing ghosting quality inspection based on a reference image, including: performing target detection or segmentation on the image to be inspected, scaling and rotating the target area so that it is aligned with the reference image; using the reference image to locate the color text area therein, and determining the candidate area of the color text area in the image to be inspected based on the correspondence between the reference image and the image to be inspected. Both the image to be inspected and the reference image are converted to the basic color space used by the printer. After the color space conversion, the template matching method is used to locate the printing position of each channel of the printed color text in the image to be inspected. By comparing whether these positions are offset, it can be determined whether there is a ghosting phenomenon.

[0004] While the aforementioned approach can effectively detect ghosting defects in printed materials using reference images, it has significant limitations in the context of archival digitization. Since the original reference images used as a comparison benchmark are often unavailable during the scanning process of historical archives, ghosting detection methods based on reference images are difficult to implement, directly impacting the integrity and reliability of archival image quality testing. Summary of the Invention

[0005] In order to solve the technical problem that the above-mentioned ghost detection method based on reference images is difficult to implement and affects the integrity and reliability of archival image quality detection, the present invention provides an archival image quality detection method, comprising:

[0006] Acquire multiple connected domains corresponding to each grayscale value in the archival image; take any grayscale value as the target grayscale value, translate the connected domain corresponding to the target grayscale value in different directions, obtain the matching rate of the target grayscale value and the remaining grayscale values in different directions and different step sizes during translation, use the direction and step size corresponding to the maximum matching rate to constitute the translation feature of the target grayscale value, and use the grayscale value corresponding to the maximum matching rate as the matching grayscale value of the target grayscale value; cluster all grayscale values into multiple categories according to the translation feature of each grayscale value; determine the grayscale concentration and matching grayscale concentration of each category according to the distribution of grayscale values in each category and the distribution of corresponding matching grayscale values; perform one-dimensional Gaussian fitting on the number of matching pixel points corresponding to the maximum matching rate of all grayscale values in each category, and obtain the fitting error according to the fitting result; determine the ghosting probability of each category according to the fitting error, grayscale concentration, matching grayscale concentration and the average of the maximum matching rate of all grayscale values in each category; determine the ghosting area in the archival image according to the size of the ghosting probability, and realize archival image quality detection.

[0007] The present invention uses multi-directional translation matching to capture the unique displacement repetitive characteristics of ghosting, thereby constructing the translation feature of the grayscale value, and classifying the grayscale values with the same or similar translation feature into one category through clustering. The ghosting probability is comprehensively evaluated by combining multi-dimensional features such as the grayscale concentration of the category, the matching grayscale concentration and the Gaussian fitting error, which can effectively distinguish between real textures and ghosting. It not only improves the detection sensitivity of slight ghosting, but also enhances the robustness of the algorithm in complex backgrounds through multi-feature fusion, providing a quantitative basis for the quality assessment of archival images, and realizing the accurate detection of ghosting areas in archival images.

[0008] Preferably, the method of obtaining multiple connected domains corresponding to each grayscale value in the archival image includes: marking the target grayscale value in the archival image as 1 and marking the remaining grayscale values as 0 to obtain a binary image corresponding to the target grayscale value, performing connected domain analysis on the binary image to obtain multiple connected domains corresponding to the target grayscale value.

[0009] Preferably, the obtaining of multiple connected domains corresponding to each grayscale value in the archival image includes: performing connectivity analysis on the target grayscale value in the archival image: when the grayscale values of two pixels in the archival image are both target grayscale values, and one of the pixels is located within the 8-neighborhood of the other pixel, the two pixels are determined to be connected to each other; all pixels that are connected to each other and whose grayscale values are the target grayscale values constitute a connected domain of the target grayscale value.

[0010] Preferably, the obtaining of the matching rate between the target grayscale value and the remaining grayscale values in different directions and different step sizes during translation includes: taking any one direction as the target direction, taking any one step size in the step size set as the target step size, translating all connected domains of the target grayscale value in the target direction by the target step size, obtaining the number of pixels that overlap with the connected domain corresponding to the 𝑘th grayscale value outside the target grayscale value after translation, and taking the ratio of the number of overlapping pixels to the number of pixels corresponding to the target grayscale value as the matching rate of the target grayscale value with the 𝑘th grayscale value outside the target grayscale value in the target direction and target step size.

[0011] The present invention can comprehensively evaluate the spatial distribution correlation between the target grayscale value and other grayscale values through multi-directional and multi-step translation matching calculations. It uses connected domains as the basic calculation unit, effectively retains the topological structure characteristics of the grayscale area in the image, and provides a quantitative basis for ghosting feature analysis.

[0012] Preferably, the step size set is obtained by: All integer values within constitute the step size set, where Indicates the length of the archive image. Indicates the width of the archive image.

[0013] Preferably, any one direction is an angle range The azimuth is divided into 360 discrete directions at intervals of 1°.

[0014] Preferably, the method for obtaining the grayscale centrality is: obtaining the mean of the differences between all adjacent grayscale values in the category, performing negative correlation mapping on the mean of the differences, normalizing the result after the negative correlation mapping, and obtaining the grayscale centrality of the category.

[0015] The present invention uses the mean of adjacent grayscale differences to capture the local aggregation characteristics of grayscale distribution, and obtains the grayscale concentration of the category by performing negative correlation mapping on the mean of the differences, providing a data basis for the subsequent acquisition of the ghosting probability of the category.

[0016] Preferably, the method for obtaining the matching grayscale centrality is: obtaining the standard deviation of the matching grayscale values corresponding to all grayscale values in the category, performing negative correlation normalization on the standard deviation, and obtaining the matching grayscale centrality of the category.

[0017] Preferably, determining the ghosting probability of each category based on the fitting error, grayscale centrality, matching grayscale centrality and the mean of the maximum matching rate of all grayscale values in each category includes: performing negative correlation normalization on the fitting error, and taking the product of the grayscale centrality, matching grayscale centrality, the mean of the maximum matching rate of all grayscale values in the category and the negative correlation normalization result of the fitting error as the ghosting probability of the category.

[0018] The present invention comprehensively considers multiple key indicators such as fitting error, grayscale concentration, matching grayscale concentration and maximum matching rate mean, uses negative correlation normalization to process the fitting error, and then performs a product operation with other parameters to finally determine the ghosting probability of the category. This not only retains the spatial characteristics of the grayscale distribution (grayscale concentration), but also takes into account the reliability of the matching results (matching rate), making the ghosting probability assessment of the category more accurate.

[0019] Preferably, the determining of the ghosting area in the archival image according to the ghosting probability includes: in response to the ghosting probability being greater than a preset probability threshold, the area formed by the pixel points corresponding to all grayscale values in the corresponding category is the ghosting area in the archival image.

[0020] The adaptive determination mechanism based on probability threshold of the present invention not only ensures the accuracy of ghost detection, but also can accurately locate the ghost area, and can flexibly adjust the detection sensitivity for archival images of different qualities.

[0021] The beneficial effects of the present invention are:

[0022] The present invention can effectively distinguish between real textures and ghosting, improving ghosting detection sensitivity. This provides a quantitative basis for quality assessment of archival images and enables accurate detection of ghosting areas in archival images. The present invention can accurately locate and detect ghosting areas without relying on the original reference image, thus having a wider range of applications and being applicable to various archival digitization scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 is a flow chart schematically illustrating a method for detecting quality of an archival image in the present invention;

[0024] Figure 2 is a schematic diagram schematically showing an archival image;

[0025] Figure 3 FIG. 1 is a diagram schematically illustrating a ghosting area. DETAILED DESCRIPTION

[0026] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work shall fall within the scope of protection of the present invention.

[0027] The specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0028] The embodiment of the present invention discloses a method for detecting the quality of an archive image. Figure 1 , including steps S1 to S7:

[0029] S1. Scan and obtain the archival image.

[0030] Specifically, during the digitization process, professional document scanning equipment is first used to perform high-precision scanning of the original archive, acquiring digital images in RGB format. To optimize the subsequent image quality testing process, the acquired RGB digital images undergo color space conversion, converting them into single-channel grayscale images.

[0031] At this point, the archival images were obtained. Figure 2 This is a schematic diagram of the archival image.

[0032] S2. Obtain multiple connected domains corresponding to each grayscale value in the archive image.

[0033] It should be noted that when scanning archives using a scanning device, if the scanning head or conveyor belt experiences jitter or uneven speed during the scanning process, ghosting may occur in the scanned archive image. The grayscale values within the ghosting region are close or similar. Therefore, the present invention first obtains multiple connected domains corresponding to each grayscale value in the archive image. The grayscale values are then classified based on the characteristics of the multiple connected domains corresponding to each grayscale value, thereby filtering out ghosting regions.

[0034] Specifically, traditional connected domain analysis is performed based on binary images. Therefore, in one embodiment, an arbitrary grayscale value is first used as the target grayscale value. The target grayscale value in the archival image is marked as 1, and the remaining grayscale values are marked as 0. This converts the archival image into a binary image corresponding to the target grayscale value. Connected domain analysis is then performed on the binary image to obtain multiple connected domains corresponding to the target grayscale value. Similarly, multiple connected domains corresponding to each grayscale value in the archival image can be obtained.

[0035] It should be noted that a connected domain refers to a set of regions consisting of interconnected pixels with the same grayscale value. Therefore, in another embodiment, connectivity analysis is performed directly on each grayscale value in the archival image: when two pixels in the archival image have the same grayscale value and one pixel is within the 8-neighborhood of the other pixel, the two pixels are considered to be interconnected, and the region formed by all interconnected pixels is a connected domain. By traversing and analyzing each grayscale value in the archival image, multiple connected domains corresponding to each grayscale value can be obtained.

[0036] S3. Take any grayscale value as the target grayscale value, translate the connected domain corresponding to the target grayscale value in different directions, obtain the matching rate between the target grayscale value and the remaining grayscale values in different directions and different step lengths during translation, and use the direction and step length corresponding to the maximum matching rate to form the translation feature of the target grayscale value. Use the grayscale value corresponding to the maximum matching rate as the matching grayscale value of the target grayscale value.

[0037] It should be noted that ghosting is repeated or blurred images in archival images. Therefore, the text in the ghosted area is repeated in the same structure elsewhere in the archival image. Therefore, the present invention performs translations of multiple connected domains corresponding to each grayscale value in different directions and step lengths. Based on the overlap between the multiple connected domains corresponding to the grayscale value after translation and the connected domains corresponding to the remaining grayscale values, the matching ratio between the grayscale value after translation and the remaining grayscale values is obtained.

[0038] Specifically, the angle range Divide the interval into 360 discrete directions at intervals of 1°. All integer values within constitute the step size set, where Indicates the length of the archive image. Indicates the width of the archival image. Set any one of the 360 discrete directions as the target direction, and any one of the step lengths set as the target step length. Implementers can set the angle range and step length set based on their specific needs.

[0039] Get the target gray value in the target direction, target step size and the target gray value. Gray value matching rate:

[0040] ;

[0041] in, Indicates the target gray value in the target direction, target step size and the target gray value. The matching rate of gray values; Indicates the number of pixels corresponding to the target grayscale value; It means that after all connected domains of the target gray value are translated in the target direction by the target step length, the first connected domains other than the target gray value in the original image are The number of pixels that overlap in the connected domain corresponding to the gray value; is the ratio of the number of overlapping pixels. When the number of overlapping pixels is larger, it means that the connected domain corresponding to the target gray value is closer to the first The shape structure of the connected domain corresponding to the gray value is similar, or the connected domain corresponding to the target gray value belongs to the first gray value after translation. The target gray value may be a part of the connected domain corresponding to the target gray value. The ghost grayscale value of the text and other image features corresponding to the grayscale value.

[0042] Furthermore, the target grayscale value and any grayscale value other than the target grayscale value have a matching rate in any direction and any step size, and the direction and step size corresponding to the maximum matching rate constitute a two-dimensional vector as the translation feature quantity of the target grayscale value, and the grayscale value corresponding to the maximum matching rate is used as the matching grayscale value of the target grayscale value.

[0043] S4. Cluster all grayscale values into multiple categories according to the translation feature of each grayscale value, and determine the grayscale concentration and matching grayscale concentration of each category according to the distribution of grayscale values in each category and the distribution of corresponding matching grayscale values.

[0044] It should be noted that each grayscale value has the highest matching rate with its corresponding matching grayscale value under its corresponding translation feature, and each grayscale value in the ghosting area can achieve the best match with the actual grayscale value of the original text in the archival image under its corresponding translation feature. Because the ghosting phenomenon generally exhibits regular spatial displacement characteristics, that is, the ghost maintains a relatively fixed distance and direction relationship with the original text, the translation feature values of different grayscale values in the ghosting area show a high degree of consistency or similarity. Therefore, the present invention divides all grayscale values into multiple categories based on the translation feature value of each grayscale value.

[0045] Specifically, the translation features of all grayscale values in the archival image are used as data samples. These data samples are clustered using the DBSCAN clustering algorithm, classifying all grayscale values in the archival image into multiple categories. Implementers can select a clustering algorithm based on actual implementation, such as mean shift clustering.

[0046] It should be noted that in the absence of noise, the grayscale values of the ghosting area are uniform. However, in the presence of noise, the grayscale values of the ghosting area are the same or similar, exhibiting the characteristics of a Gaussian distribution. Similarly, the grayscale values of the original text corresponding to the ghosting area are the same or similar. Therefore, the present invention determines the grayscale concentration and matching grayscale concentration of each category based on the distribution of grayscale values in each category and the distribution of the corresponding matching grayscale values.

[0047] Specifically, any category is taken as the target category, and the grayscale concentration of the target category satisfies the expression:

[0048] ;

[0049] Where, Indicates the grayscale concentration of the target category; Indicates the target category Grayscale values; Represents the number of grayscale values in the target category; Represents an exponential function with a natural constant as the base, used for Perform negative correlation mapping; is the mean of the differences of all adjacent gray values in the target category. When the mean of the differences is smaller, the gray value distribution in the target category is more concentrated, and the gray concentration of the target category is greater. When the mean of the differences is larger, the gray value distribution in the target category is more discrete, and the gray concentration of the target category is smaller. Since the gray values in the target category are different, the minimum mean of the differences of all adjacent gray values in the target category is 1. After negative correlation mapping, the maximum , so the present invention will Divide by , mapping the value range of grayscale centrality to the range of (0-1].

[0050] The matching grayscale concentration of the target category satisfies the expression:

[0051] ;

[0052] in, Indicates the matching grayscale concentration of the target category; Indicates the target category The matching gray value of gray values; Represents the number of grayscale values in the target category; represents the standard deviation function; Represents an exponential function with a natural constant as the base, used for Perform negative correlation normalization; when the standard deviation of the matching grayscale values of each grayscale value in the target category is smaller, it means that the matching grayscale values corresponding to each grayscale value in the target category are the same or similar, the more concentrated the distribution of the matching grayscale values of each grayscale value in the target category is, the greater the matching grayscale concentration of the target category is; when the standard deviation of the matching grayscale values of each grayscale value in the target category is larger, it means that the matching grayscale values corresponding to each grayscale value in the target category are more different, the more discrete the distribution of the matching grayscale values of each grayscale value in the target category is, the smaller the matching grayscale concentration of the target category is.

[0053] S5. Perform one-dimensional Gaussian fitting on the number of matching pixels corresponding to the maximum matching rate of all grayscale values in each category, and obtain a fitting error based on the fitting result.

[0054] Specifically, for any grayscale value in any category, the number of matching pixel points corresponding to the maximum matching rate of the grayscale value is obtained, that is, the number of pixel points that overlap with the connected domain of the matching grayscale value of the grayscale value after all connected domains of the grayscale value are translated in the direction and step size corresponding to the translation feature value.

[0055] It should be noted that, in the absence of noise interference, the grayscale values of the ghosting area are uniform, but under the interference of noise, the grayscale values of the ghosting area are the same or similar, and the overall characteristics of the Gaussian distribution are presented, that is, the number of matching pixel points corresponding to the maximum matching rate of each grayscale value in the ghosting area presents a Gaussian distribution. Therefore, the present invention performs Gaussian fitting on the number of matching pixel points corresponding to the maximum matching rate of all grayscale values in each category, so as to subsequently determine the ghosting probability of the corresponding category based on the fitting error.

[0056] Specifically, a one-dimensional Gaussian fitting is performed on the number of matching pixels corresponding to the maximum matching rate of all grayscale values in each category, and the fitting error is obtained according to the fitting result:

[0057] ;

[0058] in, represents the fitting error of the target category; Indicates the target category The number of matching pixels corresponding to the maximum matching rate of gray values; Indicates the target category Grayscale values; Indicates the target category The maximum matching rate of the gray value corresponds to the fitting value of the number of matching pixels. When the larger the value, the less the number of matching pixels corresponding to the maximum matching rate of all gray values in the target category conforms to the Gaussian distribution. On the contrary, when the fitting error is The smaller the value, the more the number of matching pixels corresponding to the maximum matching rate of all grayscale values in the target category conforms to the Gaussian distribution.

[0059] S6. Determine the ghosting probability of each category according to the fitting error, the grayscale concentration, the matching grayscale concentration, and the average of the maximum matching rates of all grayscale values in each category.

[0060] Specifically, the ghost probability of the target category satisfies the expression:

[0061] ;

[0062] in, represents the ghost probability of the target category; Represents the average of the maximum matching rates of all grayscale values in the target category; Indicates the grayscale concentration of the target category; Indicates the matching grayscale concentration of the target category; represents the fitting error of the target category; Represents an exponential function with a natural constant as the base, which is used to normalize the negative correlation of the fitting error. In archival images, ghosting areas usually account for a small proportion, while the original text area occupies a larger area. After the connected domain of the grayscale value corresponding to the original text is translated, it can only partially match the ghosting area; while after the connected domain of the grayscale value of the ghosting area is translated, it can form a better correspondence with the original text area. Therefore, the maximum matching rate of the grayscale value corresponding to the original text is small, and the maximum matching rate of the grayscale value corresponding to the ghosting area is large. When the average value of the maximum matching rate of all grayscale values in the target category is When the larger the value is, the better the effect of the connected domain of all gray values in the target category coincides with the text or pattern at another position in the archive image after translation, and the gray values in the target category are more likely to be the gray values of the ghost area; when the gray concentration of the target category and the matching gray concentration are larger, it means that the gray values in the target category are distributed similarly, and the corresponding matching gray values are also the same or similar, and the gray values in the target category are more likely to be the gray values of the ghost area; when the fitting error is larger, the gray value of the target category is closer, and the matching gray values are the same or similar, and the gray values in the target category are more likely to be the gray values of the ghost area; The smaller it is, the more the number of matching pixels corresponding to the maximum matching rate of all grayscale values in the target category conforms to the Gaussian distribution, and the grayscale values in the target category are more likely to be the grayscale values of the ghost area.

[0063] S7. Determine the ghosting area in the archive image according to the ghosting probability to achieve archive image quality detection.

[0064] Specifically, in response to the ghost probability being greater than a preset probability threshold, the area formed by the pixels corresponding to all grayscale values in the corresponding category is the ghost area in the archive image. For a schematic diagram of the ghost area, see Figure 3 .

[0065] The probability threshold is set by the implementer based on the actual implementation situation, for example, 0.3.

[0066] At this point, archival image quality detection has been achieved.

Claims

1. A method for detecting the quality of an archive image, characterized in that: include: Obtain multiple connected domains corresponding to each grayscale value in the archival image; Take any grayscale value as the target grayscale value, translate the connected domain corresponding to the target grayscale value in different directions, and obtain the matching rate between the target grayscale value and the other grayscale values in different directions and different step lengths during translation, including: Take any direction as the target direction, any step in the step set as the target step, translate all connected domains of the target gray value in the target direction by the target step, and obtain all connected domains of the target gray value after translation and the first connected domain outside the target gray value. The number of pixels that overlap in the connected domain corresponding to the target gray value is calculated, and the ratio of the number of overlapping pixels to the number of pixels corresponding to the target gray value is used as the first pixel value outside the target gray value in the target direction and target step size. The matching rate of gray values; The direction and step length corresponding to the maximum matching rate constitute the translation feature of the target grayscale value, and the grayscale value corresponding to the maximum matching rate is used as the matching grayscale value of the target grayscale value; Clustering all grayscale values into multiple categories according to the translation feature of each grayscale value; Determining the grayscale concentration and matching grayscale concentration of each category based on the distribution of grayscale values in each category and the distribution of corresponding matching grayscale values; performing a one-dimensional Gaussian fit on the number of matching pixels corresponding to the maximum matching rate of all grayscale values in each category, and obtaining a fitting error based on the fitting result; and determining the ghosting probability of each category based on the fitting error, the grayscale concentration, the matching grayscale concentration, and the average of the maximum matching rate of all grayscale values in each category; The ghosting area in the archival image is determined according to the ghosting probability, so as to realize the archival image quality detection.

2. The method for detecting the quality of an archive image according to claim 1, wherein: The step of obtaining multiple connected domains corresponding to each grayscale value in the archive image includes: The target grayscale value in the archive image is marked as 1, and the remaining grayscale values are marked as 0 to obtain a binary image corresponding to the target grayscale value. Connected domain analysis is performed on the binary image to obtain multiple connected domains corresponding to the target grayscale value.

3. The method for detecting the quality of an archive image according to claim 1, wherein: The step of obtaining multiple connected domains corresponding to each grayscale value in the archive image includes: Connectivity analysis of the target grayscale value in the archival image is performed: when the grayscale values of two pixels in the archival image are both the target grayscale value, and one of the pixels is located in the 8-neighborhood of the other pixel, the two pixels are determined to be connected to each other; all pixels that are connected to each other and have grayscale values equal to the target grayscale value constitute a connected domain of the target grayscale value.

4. The method for detecting the quality of an archive image according to claim 1, wherein: The method for obtaining the step size set is: The interval All integer values within constitute the step size set, where Indicates the length of the archive image. Indicates the width of the archive image.

5. The method for detecting the quality of an archive image according to claim 1, wherein: Any one direction is an angle range The azimuth is divided into 360 discrete directions at intervals of 1°.

6. The method for detecting the quality of an archive image according to claim 1, wherein: The method for obtaining the grayscale centrality is: The mean of the differences between all adjacent grayscale values in the category is obtained, negative correlation mapping is performed on the mean of the differences, and the result after the negative correlation mapping is normalized to obtain the grayscale concentration of the category.

7. The method for detecting the quality of an archive image according to claim 1, wherein: The method for obtaining the matching grayscale centrality is: The standard deviation of the matching grayscale values corresponding to all grayscale values in the category is obtained, and the standard deviation is negatively normalized to obtain the matching grayscale concentration of the category.

8. The method for detecting the quality of an archive image according to claim 1, wherein: Determining the ghosting probability of each category based on the fitting error, grayscale concentration, matching grayscale concentration, and the average of the maximum matching rates of all grayscale values in each category includes: The fitting error is negatively normalized; and the product of the grayscale centrality, the matching grayscale centrality, the mean of the maximum matching rate of all grayscale values in the category, and the negative correlation normalization result of the fitting error is used as the ghosting probability of the category.

9. The method for detecting the quality of an archive image according to claim 1, wherein: Determining the ghost area in the archive image according to the ghost probability includes: In response to the ghost probability being greater than a preset probability threshold, an area formed by pixels corresponding to all grayscale values in the corresponding category is a ghost area in the archive image.

Images