Archival picture intelligent noise reduction method

Abstract

The present application relates to the field of image processing, and more particularly to an intelligent noise reduction method for archive pictures, comprising: determining an optimization analysis strategy for each target processing image according to an abnormality proportion index and an abnormality distribution index, and performing auxiliary feature analysis or abnormality radiation analysis on the target processing image; when performing auxiliary feature analysis, determining a setting mode of a feature analysis set of each target processing image of a first type and an auxiliary analysis mode according to a correlation richness coefficient; when performing abnormality radiation analysis, determining a noise reduction execution mode of each abnormal point in each target processing image of a second type according to a processing radiation coefficient, and determining a noise reduction execution index of the abnormal point based on the processing radiation coefficient and a region radiation coefficient, or a region radiation proportion and a region abnormality parameter; and performing noise reduction processing on each target processing image based on the determined noise reduction execution index, thereby improving the image quality of the target processing image after noise reduction processing.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of image processing, and in particular to an intelligent noise reduction method for archive pictures. BACKGROUND

[0002] Through image scanning of the stored paper archives to complete the electronic management of the archives, the efficiency of the archive management can be ensured while avoiding the loss or damage of paper archives affecting the integrity of the archive information. However, in the actual process of image scanning of paper archives, various paper defects will inevitably be encountered, and noise reduction processing needs to be performed on the archive images to ensure the effectiveness of the obtained archive image information. However, in the existing noise reduction processing of paper archive images, the targeted noise reduction processing means is not considered for different defect characteristics, resulting in poor effectiveness of the obtained archive image information. Therefore, how to determine a targeted noise reduction processing method according to the actual obtained archive image to ensure the effectiveness of the archive image information is a problem to be solved by those skilled in the art.

[0003] Chinese Patent Application Publication No. CN113610715A discloses a digital archive image noise reduction processing method, relating to the technical field of digital archive images. Each frame of image data is obtained, the obtained image data is analyzed for noise, the difference frame image between the image data is calculated, the noise estimation parameter of the image data is determined according to the difference frame image of the image data, the noise estimation parameter is used to perform noise estimation processing on each frame of image data, and the noise level information of each frame of image data is determined. The noise type and noise power of the current frame are calculated based on the divided basic block, the noise reduction parameter is adjusted based on the calculated noise type and noise power to retain the main characteristics of the image data, the image data determined as noise is processed based on the noise reduction parameter, and the image data after noise reduction is determined, which can effectively reduce the noise of the image data. However, the above-mentioned scheme has the following defects: the defect characteristic analysis based on the obtained paper archive image is not performed to determine the targeted noise reduction processing means, resulting in the risk of excessive noise reduction processing of the archive image, and thus the effectiveness of the obtained archive image information is low. SUMMARY

[0004] Therefore, the present application provides an intelligent noise reduction method for archive pictures to overcome the problem that the prior art fails to perform defect characteristic analysis based on the obtained paper archive image to determine the targeted noise reduction processing means, resulting in the risk of excessive noise reduction processing of the archive image, and thus the effectiveness of the obtained archive image information is low.

[0005] To achieve the above-mentioned purpose, the present application provides an intelligent noise reduction method for archive pictures, comprising:

[0006] The optimization analysis strategy of each target processing image is determined according to the abnormal proportion index and the abnormal distribution index, and the optimization analysis strategy is auxiliary feature analysis or abnormal radiation analysis on the target processing image;

[0007] In the auxiliary feature analysis, the setting mode of the feature analysis set of each target processing image of a category and the auxiliary analysis mode are determined according to the correlation richness coefficient, the setting mode of the feature analysis set is determined based on the feature distribution correlation coefficient or the storage parameter correlation degree to determine the feature analysis set of the target processing image of the category, and the auxiliary analysis mode is determined based on the set reference key index or the defect confidence degree to determine the noise reduction processing mode to obtain the denoising execution index of each abnormal sub-region;

[0008] In the abnormal radiation analysis, the denoising execution mode of each abnormal point in each target processing image of a category is determined according to the processing radiation coefficient, and the denoising execution index of the abnormal point is determined based on the processing radiation coefficient and the region radiation coefficient, or the region radiation proportion and the region abnormal parameter;

[0009] Based on the determined denoising execution index, the denoising processing is performed on each target processing image.

[0010] Further, if there is a target processing image whose abnormal proportion index is greater than a preset abnormal proportion index or whose abnormal distribution index is greater than a preset abnormal distribution index, the optimization analysis strategy is auxiliary feature analysis on the target processing image;

[0011] The target processing image whose abnormal proportion index is greater than a preset abnormal proportion index or whose abnormal distribution index is greater than a preset abnormal distribution index is recorded as a target processing image of a category.

[0012] Further, if there is a target processing image whose abnormal proportion index is less than or equal to a preset abnormal proportion index and whose abnormal distribution index is less than or equal to a preset abnormal distribution index, the optimization analysis strategy is abnormal radiation analysis on the target processing image;

[0013] The target processing image whose abnormal proportion index is less than or equal to a preset abnormal proportion index and whose abnormal distribution index is less than or equal to a preset abnormal distribution index is recorded as a target processing image of a category.

[0014] Further, if there is a target processing image of a category whose correlation richness coefficient is greater than a preset correlation richness coefficient, the feature analysis set is determined based on the feature distribution correlation coefficient;

[0015] The set reference key index of each abnormal sub-region is determined according to the abnormal reference index and the reference gradient index, and the noise reduction processing mode of each abnormal sub-region is determined based on the set reference key index.

[0016] Further, if the set reference key index of the abnormal sub-region is greater than the preset set reference key index, the denoising execution index of the abnormal sub-region is determined according to the set reference key index and the set reference index;

[0017] If the set reference key index of the abnormal sub-region is less than or equal to the preset set reference key index, the denoising execution index of the abnormal sub-region is determined according to the reference gradient index.

[0018] Further, if the association richness coefficient of a target processing image of a class is less than or equal to the preset association richness coefficient, the feature analysis set of the target processing image of the class is determined based on the storage parameter correlation degree;

[0019] The defect confidence degree is determined according to the edge index difference value and the set verification index, and the denoising processing mode of each abnormal sub-region is determined based on the defect confidence degree.

[0020] Further, if the defect confidence degree of a target processing image of a class is greater than the preset defect confidence degree, the denoising execution index of each abnormal sub-region is determined according to the defect edge parameter and the defect confidence degree;

[0021] If the defect confidence degree of a target processing image of a class is less than or equal to the preset defect confidence degree, the denoising execution index of each abnormal sub-region is determined according to the reference gradient index and the defect confidence degree.

[0022] Further, under the abnormal radiation analysis condition, the processing radiation coefficient of each abnormal point in the target processing image of the second class is determined according to the region gradient index and the gradient key index, and the denoising execution mode of each abnormal point is determined according to the processing radiation coefficient;

[0023] The processing radiation coefficient is positively correlated with the region gradient index and the gradient key index, respectively;

[0024] The abnormal radiation analysis condition is that the target processing image is optimized and analyzed by abnormal radiation analysis.

[0025] Further, for a single target processing image of the second class, if the processing radiation coefficient of an abnormal point is greater than the preset processing radiation coefficient, the denoising execution index of the abnormal point is determined based on the processing radiation coefficient and the region radiation coefficient;

[0026] The denoising execution index is positively correlated with the processing radiation coefficient, and the denoising execution index is negatively correlated with the region radiation coefficient.

[0027] Further, for a single target processing image of the second class, if the processing radiation coefficient of an abnormal point is less than or equal to the preset processing radiation coefficient, the denoising execution index of the abnormal point is determined based on the region radiation proportion and the region abnormal parameter.

[0028] The denoising execution index is negatively correlated with the area radiation proportion and the area anomaly parameter, respectively.

[0029] Compared with the prior art, the beneficial effects of the present application are that the present application determines the optimization analysis strategy of each target processing image according to the anomaly proportion index and the anomaly distribution index to complete the denoising process by adopting a targeted optimization analysis strategy for the target processing image, so as to ensure that the denoising process of the target processing image is more in line with the actual situation of the target processing image, and the present application avoids the over-processing of the denoising process of the archive image, which leads to poor effectiveness of the target processing image.

[0030] Further, in the present application, the interference degree of the suspected defect points in the target processing image is represented by the anomaly proportion index and the anomaly distribution index, and a targeted optimization analysis strategy is adopted for target processing images with different interference degrees, for example, in the case of a larger anomaly proportion index or a larger anomaly distribution index, the evaluation of the existence of different regions of text is greatly affected, and the above-mentioned influence is compensated by auxiliary feature analysis, on the contrary, in the case of smaller anomaly proportion index and smaller anomaly distribution index, the target processing image is less affected, and the determination of the text coverage area is not disturbed, therefore, the influence of each anomaly point on the text area during the denoising process is determined by the abnormal radiation analysis, which ensures the analysis efficiency of the denoising process of the target processing image while ensuring the denoising effect of the target processing image.

[0031] Further, in the present application, for a type of target processing image, the setting mode and auxiliary analysis mode of the feature analysis set of the type of target processing image are determined according to the correlation richness coefficient, so that the division process of the determined feature analysis set is more in line with the actual situation, while ensuring the auxiliary effect and division efficiency of the feature analysis set, and corresponding targeted denoising process is performed, which ensures the denoising effect of the target processing image while avoiding interference caused by text recognition during the denoising process of the target processing image.

[0032] Further, in the present application, for a type of target processing image, the setting mode and auxiliary analysis mode of the feature analysis set of the type of target processing image are determined according to the correlation richness coefficient, so that the division process of the determined feature analysis set is more in line with the actual situation, while ensuring the auxiliary effect and division efficiency of the feature analysis set, and corresponding targeted denoising process is performed, which ensures the denoising effect of the target processing image while avoiding interference caused by text recognition during the denoising process of the target processing image. BRIEF DESCRIPTION OF DRAWINGS

[0033] Fig. 1 The schematic diagram of the file picture intelligent noise reduction method of the present application is shown in FIG. 1.

[0034] Fig. 2 The flow chart of determining the optimization analysis strategy according to the abnormal proportion index and the abnormal distribution index of the present application is shown in FIG. 2.

[0035] Fig. 3 The flow chart of determining the auxiliary analysis mode of the first type of target processing image according to the correlation richness coefficient of the present application is shown in FIG. 3.

[0036] Fig. 4 The flow chart of determining the noise reduction execution mode of each abnormal point in the second type of target processing image according to the processing radiation coefficient of the present application is shown in FIG. 4. DETAILED DESCRIPTION

[0037] In order to make the objects and advantages of the present application clearer, the present application will be further described below in conjunction with embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and do not limit the present application.

[0038] The preferred embodiments of the present application will be described below with reference to the accompanying drawings. It should be understood by those skilled in the art that these embodiments are only used to explain the technical principles of the present application and are not intended to limit the protection scope of the present application.

[0039] It should be noted that in the description of the present application, the terms of direction or position relationship such as "upper", "lower", "left", "right", "inner", "outer" and the like are based on the direction or position relationship shown in the drawings, which is only for the convenience of description and does not indicate or imply that the device or element must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present application.

[0040] In addition, it should also be noted that in the description of the present application, unless otherwise explicitly specified and limited, the terms "mounting", "connecting", "connection" should be understood in a broad sense, for example, it can be fixed connection, or detachable connection, or integrally connected; it can be mechanical connection, or electrical connection; it can be directly connected, or indirectly connected through an intermediate medium, or the internal communication of two elements. Those skilled in the art can understand the specific meaning of the above terms in the present application according to the specific circumstances.

[0041] Please refer to Figs. 1 to 4 The present application provides a file picture intelligent noise reduction method, which comprises:

[0042] Determining the optimization analysis strategy of each target processing image according to the abnormal proportion index and the abnormal distribution index is to perform auxiliary feature analysis or abnormal radiation analysis on the target processing image.

[0043] In the auxiliary feature analysis, the setting mode of the feature analysis set of each target processing image of one type of target is determined according to the correlation richness coefficient, and the auxiliary analysis mode is determined according to the feature distribution correlation coefficient or the storage parameter correlation degree, and the auxiliary analysis mode is determined according to the set reference key index or the defect confidence degree, and the noise reduction processing mode is determined, so as to obtain the denoising execution index of each abnormal sub-region;

[0044] In the abnormal radiation analysis, the denoising execution mode of each abnormal point in each target processing image of two types of targets is determined according to the processing radiation coefficient, and the denoising execution index of the abnormal point is determined according to the processing radiation coefficient and the region radiation coefficient, or the region radiation proportion and the region abnormal parameter;

[0045] Based on the determined denoising execution index, the denoising processing is performed on each target processing image.

[0046] In the present application, the noise reduction processing is performed on the obtained archive image in the scanning process of the paper archive, so as to ensure the integrity of the obtained archive image information, and the archive to be scanned and uploaded is recorded as a target processing archive, and the image obtained by scanning each information page contained in the target processing archive is recorded as a target processing image, the information page is a page with text or image information in the paper material corresponding to the target processing archive, and the specification of the target processing image obtained in the present application is consistent;

[0047] In the present application, a plurality of image optimization records are applied, and each image optimization record records at least one abnormal coefficient, abnormal proportion index, correlation richness coefficient, feature distribution correlation coefficient, distribution correlation proportion, set reference key index, storage parameter correlation degree, defect confidence degree, denoising execution index and processing radiation coefficient in the process of optimizing analysis of the target processing image, and each image optimization record corresponds to a qualified mark, and the qualified mark records whether the effectiveness of the target processing image after completing the noise reduction processing meets the user's demand, and it can be understood that the user can determine whether the effectiveness of the target processing image after completing the noise reduction processing meets the demand according to the self-set index, for example, the self-set index can be but is not limited to the recognition obstacle index, and the recognition obstacle index = the number of target processing images with text recognition failure after completing the optimization analysis / the number of target processing images after completing the optimization analysis.

[0048] Specifically, if the abnormal proportion index of the target processing image is greater than the preset abnormal proportion index or the abnormal distribution index is greater than the preset abnormal distribution index, the optimization analysis strategy is auxiliary feature analysis for the target processing image;

[0049] The target processing image with the abnormal proportion index greater than the preset abnormal proportion index or the abnormal distribution index greater than the preset abnormal distribution index is recorded as a first type of target processing image.

[0050] Specifically, if the abnormal proportion index of the target processing image is less than or equal to the preset abnormal proportion index and the abnormal distribution index is less than or equal to the preset abnormal distribution index, the optimization analysis strategy is to perform abnormal radiation analysis on the target processing image.

[0051] The target processing image with the abnormal proportion index less than or equal to the preset abnormal proportion index and the abnormal distribution index less than or equal to the preset abnormal distribution index is recorded as a second type of target processing image.

[0052] For a single target processing image, a plurality of to-be-analyzed points are determined, the number of pixel points included in each to-be-analyzed point is the same, the user can set the number of pixel points included in each to-be-analyzed point according to the actual working scene, the abnormal coefficient of each to-be-analyzed point in the target processing image is detected, and the to-be-analyzed point with the abnormal coefficient greater than the preset abnormal coefficient is recorded as an abnormal point. For a single to-be-analyzed point, the abnormal coefficient is determined according to the edge pixel proportion and the color channel difference parameter of the to-be-analyzed point. The abnormal coefficient is the product of the edge pixel proportion and the color channel difference parameter of the to-be-analyzed point. The edge pixel points of the pixel region corresponding to the to-be-analyzed point are detected by Canny operator detection. The edge pixel proportion = the number of edge pixel points of the pixel region corresponding to the to-be-analyzed point / the number of pixel points in the pixel region corresponding to the to-be-analyzed point. The number of pixel points of the RGB three color channels is detected respectively. The color channel difference parameter is the difference between the maximum and minimum number of pixel points included in three color channels.

[0053] The user can determine the value of the preset abnormal coefficient according to the actual working scene. For example, the user can set it according to the image optimization record. The higher the user's requirement for the effectiveness of the target processing image after noise reduction processing, the smaller the value of the preset abnormal coefficient. A method for determining the value of the preset abnormal coefficient is provided. The average value of the abnormal coefficients of each abnormal point in the image optimization record that meets the user's requirement for the effectiveness of the target processing image after noise reduction processing is recorded as the preset abnormal coefficient.

[0054] The abnormality proportion index of the target processing image is the number of abnormal points contained in the target processing image / the number of points to be analyzed contained in the target processing image, the target processing image is uniformly divided into a plurality of rectangular regions with the same area, the rectangular regions obtained by the division are referred to as sub-image regions, the specifications of each target processing image obtained in the application are consistent, and the sub-image region division results are also consistent, the sub-image regions with abnormal points are referred to as abnormal sub-regions, and the abnormal distribution index is the number of abnormal sub-regions existing in the target processing image / the number of sub-image regions existing in the target processing image.

[0055] For a single target processing image, if the abnormality proportion index of the target processing image is greater than the preset abnormality proportion index or the abnormal distribution index is greater than the preset abnormal distribution index, auxiliary feature analysis is performed on the target processing image, and the target processing image is referred to as a first type of target processing image. This type of target processing image often has more abnormal conditions or larger abnormal ranges, and the existing defect features are easy to affect the judgment result of the text coverage area. Therefore, the probability that each abnormal sub-region may affect the recognition of the text content when performing noise processing is determined by determining the feature analysis set. If the abnormality proportion index of the target processing image is less than or equal to the preset abnormality proportion index and the abnormal distribution index is less than or equal to the preset abnormal distribution index, abnormal radiation analysis is performed on the target processing image, and the target processing image is referred to as a second type of target processing image. The abnormal conditions of this type of target processing image are less, and the determination process of the text coverage area has less influence. The existence of the text in each region can be accurately determined, and the specific processing process for each abnormal point is determined through abnormal radiation analysis. While ensuring the noise reduction effect, the influence degree on the text area in the noise reduction process is effectively avoided.

[0056] The values of the preset abnormality proportion index and the preset abnormal distribution index can be determined by the user according to the actual working scene. For example, the user can set according to the image optimization record. The higher the user's requirement for the effectiveness of the target processing image after completing the noise reduction processing, the smaller the value of the preset abnormality proportion index, and the smaller the value of the preset abnormal distribution index. A method for determining the value of the preset abnormality proportion index is provided. The image optimization record subjected to abnormal radiation analysis of the target processing image is referred to as an analysis reference record. The maximum value of the abnormality proportion index in the analysis reference record that meets the user's requirement for the effectiveness of the target processing image after completing the noise reduction processing is referred to as the preset abnormality proportion index. A method for determining the value of the preset abnormal distribution index is provided. The maximum value of the abnormal distribution index in the analysis reference record that meets the user's requirement for the effectiveness of the target processing image after completing the noise reduction processing is referred to as the preset abnormal distribution index.

[0057] Specifically, if there is a type of target processing image whose correlation richness coefficient is greater than a preset correlation richness coefficient, a feature analysis set is determined based on the feature distribution correlation coefficient;

[0058] The set reference key index of each abnormal sub-region is determined according to the abnormal reference index and the reference gradient index, and the noise reduction processing mode of each abnormal sub-region is determined based on the set reference key index.

[0059] The correlation richness coefficient is determined according to the page richness index and the one-type page proportion for a single one-type target processing image, the correlation richness coefficient = ln (page richness index * one-type page proportion), the page richness index is the number of information pages existing in the target processing file to which the one-type target processing image belongs, and the one-type page proportion = the number of one-type target processing images corresponding to the information pages existing in the target processing file to which the one-type target processing image belongs / the number of information pages existing in the target processing file to which the one-type target processing image belongs.

[0060] For a single one-type target processing image, if the correlation richness coefficient of the one-type target processing image is greater than a preset correlation richness coefficient, it indicates that the one-type target processing image exists under the same storage condition and can be used to analyze the target processing image combined for analysis, and thus the specific noise reduction processing mode is determined by combining analysis with each target processing image in the target processing file to which it belongs. The value of the preset correlation richness coefficient can be determined by the user according to the actual working scene. For example, the user can set it according to the image optimization record. The higher the user's requirement for the effectiveness of the target processing image after noise reduction processing, the value of the preset correlation richness coefficient. A method for providing a value of the preset correlation richness coefficient is provided. The image optimization record for determining the feature analysis set of the one-type target processing image based on the feature distribution correlation coefficient is recorded as a one-type analysis reference record. The minimum value of the correlation richness coefficient in the one-type analysis reference record that meets the user's requirement for the effectiveness of the target processing image after noise reduction processing is recorded as the preset correlation richness coefficient.

[0061] For a single correlation rich coefficient greater than a preset correlation rich coefficient, a target processing image of a type is obtained, a feature analysis set of the target processing image of the type is obtained, the feature analysis set only contains part of the target processing images of the type corresponding to the target processing archives, and the feature distribution correlation coefficient of each target processing image in the feature analysis set is greater than a preset feature distribution correlation coefficient. For a single target processing image, the feature distribution correlation coefficient = the number of feature distribution correlation regions / the number of sub-image regions included in the target processing image. Coordinates are set for each sub-image region, for example, for a single sub-image region, if its position is in the i-th column from left to right and the j-th row from top to bottom in the division result of the target processing image in which it exists, the coordinates of the sub-image region are (i, j). For a single abnormal sub-region, if the distribution correlation proportion of the abnormal sub-region is greater than a preset distribution correlation proportion, the abnormal sub-region is recorded as a feature distribution correlation region. The distribution correlation proportion = the number of abnormal sub-regions with the same coordinates as the abnormal sub-region in the feature analysis set / the number of target processing images in the feature analysis set.

[0062] The values of the preset distribution correlation coefficient and the preset distribution correlation proportion can be determined by the user according to the actual working scenario. For example, the user can set them according to the image optimization record. The higher the user's requirement for the effectiveness of the target processing image after noise reduction, the greater the value of the preset distribution correlation coefficient and the preset distribution correlation proportion. A method for determining the value of the preset distribution correlation coefficient is provided. The average value of the feature distribution correlation coefficients of each target processing image in an analysis reference record that meets the user's requirement for the effectiveness of the target processing image after noise reduction is recorded as the preset feature distribution correlation coefficient. A method for determining the value of the preset distribution correlation proportion is provided. The average value of the distribution correlation proportions of each feature distribution correlation region in an analysis reference record that meets the user's requirement for the effectiveness of the target processing image after noise reduction is recorded as the preset distribution correlation proportion.

[0063] For a single target processing image of a type, the set reference key index is detected for each sub-region image set when the feature analysis set is determined. For any abnormal sub-region, the set reference key index = ln (abnormal reference index x reference gradient index), the abnormal reference index is the number of abnormal sub-regions in the sub-image region corresponding to the coordinates of the abnormal sub-region in the determined feature analysis set, and the reference gradient index is the average value of the gradient intensity values obtained by gradient intensity detection on the sub-image region corresponding to the coordinates of the abnormal sub-region in the feature analysis set.

[0064] Specifically, if the set reference key index of the abnormal sub-region is greater than the preset set reference key index, the denoising execution index of the abnormal sub-region is determined according to the set reference key index and the set reference index;

[0065] If the set reference key index of the abnormal sub-region is less than or equal to the preset set reference key index, the denoising execution index of the abnormal sub-region is determined according to the reference gradient index.

[0066] The value of the preset set reference key index can be determined by the user according to the actual working scene. For example, the user can set it according to the image optimization record. The higher the user's requirement for the effectiveness of the target processing image after completing the denoising processing, the smaller the value of the preset set reference key index. A method for determining the value of the preset set reference key index is provided. The image optimization record in which the denoising execution index of the sub-image region is determined according to the set reference key index and the set reference index is recorded as the key reference record. The minimum value of the set reference key index in the key reference record that meets the user's requirement for the effectiveness of the target processing image after completing the denoising processing is recorded as the preset set reference key index.

[0067] For a single abnormal sub-region, if the set reference key index is greater than the preset set reference key index, the denoising execution index and the region key coefficient are in a negative correlation relationship. The region key coefficient is the product of the set reference key index and the set reference index. The set reference index is the average value of the feature distribution correlation coefficients of each target processing image in the feature analysis set corresponding to the abnormal sub-region. If the set reference key index is less than or equal to the preset set reference key index, the denoising execution index and the reference gradient index are in a negative correlation relationship.

[0068] Specifically, if the association richness coefficient of a type of target processing image is less than or equal to the preset association richness coefficient, the feature analysis set of the type of target processing image is determined based on the storage parameter correlation degree.

[0069] The defect confidence degree is determined according to the edge index difference value and the set verification index, and the denoising processing mode of each abnormal sub-region is determined based on the defect confidence degree.

[0070] The feature analysis set of the one type of target processing image is obtained, the storage parameter correlation degree of any one target processing image in the determined feature analysis set is greater than the preset storage parameter correlation degree, and all are one type of target processing image, for a single target processing image, the storage parameter correlation degree = 1 / sum of parameter differences of each storage parameter between the target processing image and the paper material corresponding to the one type of target processing image, for a single storage parameter, the parameter difference = absolute value of the difference between the numerical value of the storage parameter corresponding to the target processing image and the numerical value of the storage parameter corresponding to the one type of target processing image / the numerical value of the storage parameter corresponding to the one type of target processing image, the category composition of the storage parameter for determining the storage parameter correlation degree is provided, the category of the storage parameter includes: storage reference temperature, storage reference humidity, storage time length and storage exposure frequency, the storage reference temperature is the average temperature of the storage process of the target processing file to which the target processing image belongs, the storage reference humidity is the average humidity of the storage process of the target processing file to which the target processing image belongs, the storage time length is the time length corresponding to the storage process of the target processing file to which the target processing image belongs, and the storage exposure frequency is the number of times of viewing the storage process of the target processing file to which the target processing image belongs.

[0071] The preset storage parameter correlation degree can be determined by the user according to the actual working scene, for example, the user can set it according to the image optimization record, the higher the requirement of the user for the effectiveness of the target processing image after noise reduction processing, the greater the value of the preset storage parameter correlation degree, and a method for determining the value of the preset storage parameter correlation degree is provided, the average value of the storage parameter correlation degrees of each target processing image in the feature analysis set in the one type of analysis reference record meeting the requirement of the user for the effectiveness of the target processing image after noise reduction processing is taken as the preset storage parameter correlation degree.

[0072] For a single one type of target processing image, the defect confidence degree = set verification index / edge index difference value, the set verification index is the average value of the storage parameter correlation degrees of each target processing image in the feature analysis set, and the edge index difference value m is the number of target processing images in the feature analysis set, dj is the edge index of the jth target processing image in the feature analysis set, d0 is the average value of the edge indexes of each target processing image in the feature analysis set, and the edge index is the maximum value of the shortest vertical distance between the center of gravity of each abnormal sub-region in the target processing image and the edge of the target processing image.

[0073] Specifically, if the defect confidence degree of a target processing image of the first category is greater than the preset defect confidence degree, the denoising execution index of each abnormal sub-region is determined according to the defect edge parameter and the defect confidence degree;

[0074] If the defect confidence degree of a target processing image of the first category is less than or equal to the preset defect confidence degree, the denoising execution index of each abnormal sub-region is determined according to the reference gradient index and the defect confidence degree.

[0075] The value of the preset defect confidence degree can be determined by the user according to the actual working scene. For example, the user can set it according to the image optimization record. The higher the user's requirement for the effectiveness of the target processing image after denoising processing, the greater the value of the preset defect confidence degree. A method for determining the value of the preset defect confidence degree is provided. The image optimization record for determining the denoising execution index of each abnormal sub-region according to the defect edge parameter and the defect confidence degree is recorded as a defect reference record. The minimum value of the defect confidence degree in the defect reference record that meets the user's requirement for the effectiveness of the target processing image after denoising processing is recorded as the preset defect confidence degree.

[0076] For any abnormal sub-region in a single target processing image of the first category, if the defect confidence degree is greater than the preset defect confidence degree, the denoising execution index has a positive correlation with the region defect coefficient, and the region defect parameter is the product of the defect edge parameter and the defect confidence degree. The defect edge parameter is the shortest vertical distance between the center of gravity of the abnormal sub-region and the edge of the target processing image. If the defect confidence degree is less than or equal to the preset defect confidence degree, the denoising execution index has a negative correlation with the analysis reliability coefficient, and the analysis reliability coefficient is the product of the reference gradient index and the defect confidence degree of the abnormal sub-region.

[0077] Specifically, under the abnormal radiation analysis condition, the processing radiation coefficient of each abnormal point in the target processing image of the second category is determined according to the region gradient index and the gradient key index, and the denoising execution mode of each abnormal point is determined according to the processing radiation coefficient.

[0078] The processing radiation coefficient has a positive correlation with the region gradient index and the gradient key index, respectively.

[0079] The abnormal radiation analysis condition is that the target processing image is optimized and analyzed by abnormal radiation analysis.

[0080] Wherein, for a single second-class target processing image, the processing radiation coefficient of any abnormal point in the image is = ln (region gradient index x gradient key index), the gradient intensity of the image region corresponding to the gradient evaluation range of the abnormal point is detected, and the obtained gradient intensity value is recorded as the region gradient index of the abnormal point. The gradient evaluation range is a square region, and the center of the circumscribed circle of the square region is the barycenter of the pixel region corresponding to the abnormal point. The number of points involved in the analysis in the horizontal and vertical directions of the square region is the same. The number of points involved in the analysis in the horizontal and vertical directions and the abnormal coefficient of the abnormal point are positively correlated. The horizontal direction is perpendicular to the vertical edge of the target processing image, and the vertical direction is perpendicular to the horizontal edge of the target processing image. The gradient key index = (the region gradient index of the abnormal point - the average value of the region gradient index of each abnormal point in the second-class target processing image) / the average value of the region gradient index of each abnormal point in the second-class target processing image.

[0081] Under the condition of abnormal radiation analysis, there are usually only a small number of abnormal points in the second-class target processing image. Therefore, in the case of ensuring the analysis efficiency of the noise reduction process of the second-class target processing image, the analysis of the text involved in the region of each abnormal point is carried out to set the denoising execution index of the noise reduction process of different abnormal points, so as to avoid the denoising process of the abnormal point involving more text regions being too strong, which may cause misprocessing of the text content.

[0082] Specifically, for a single second-class target processing image, if the processing radiation coefficient of the abnormal point is greater than the preset processing radiation coefficient, the denoising execution index of the abnormal point is determined based on the processing radiation coefficient and the region radiation coefficient.

[0083] The denoising execution index and the processing radiation coefficient are positively correlated, and the denoising execution index and the region radiation coefficient are negatively correlated.

[0084] Specifically, for a single second-class target processing image, if the processing radiation coefficient of the abnormal point is less than or equal to the preset processing radiation coefficient, the denoising execution index of the abnormal point is determined based on the region radiation proportion and the region abnormal parameter.

[0085] The denoising execution index and the region radiation proportion and the region abnormal parameter are negatively correlated.

[0086] The preset processing radiation coefficient value can be determined by the user according to an actual working scenario. For example, the user can set the preset processing radiation coefficient value according to image optimization records. The higher the user's requirement for the effectiveness of the target processing image after noise reduction processing, the greater the preset processing radiation coefficient value. A method for determining the preset processing radiation coefficient value is provided. An image optimization record for determining the de-noising execution index of an abnormal point based on the processing radiation coefficient and the area radiation coefficient is recorded as a radiation reference record. The average value of the processing radiation coefficient of each abnormal point in the radiation reference record that meets the user's requirement for the effectiveness of the target processing image after noise reduction processing is recorded as the preset processing radiation coefficient.

[0087] For a single abnormal point, if the processing radiation coefficient of the abnormal point is greater than the preset processing radiation coefficient, the de-noising execution index and the radiation reference coefficient are in a positive correlation. The radiation reference coefficient = the processing radiation coefficient of the abnormal point / area radiation coefficient. The area radiation coefficient is the sum of the product of the area abnormal parameter and the area reference radiation coefficient and the corresponding evaluation coefficient. The area abnormal parameter is the number of abnormal points in the reference evaluation range of the abnormal point. The area reference radiation coefficient is the average value of the processing radiation coefficient of each abnormal point in the reference evaluation range of the abnormal point. The reference evaluation range is a circular area with the center of the corresponding pixel area as the center and the evaluation length as the radius. The evaluation length and the processing radiation coefficient of the abnormal point are in a positive correlation. The user can set the evaluation coefficient corresponding to the area abnormal parameter and the area reference radiation coefficient according to the actual working scenario. The evaluation coefficient corresponding to the area abnormal parameter is 0.3, and the evaluation coefficient corresponding to the area reference radiation coefficient is 0.7. If the processing radiation coefficient of the abnormal point is less than or equal to the preset processing radiation coefficient, the de-noising execution index and the area key coefficient are in a negative correlation. The area key coefficient is the product of the area radiation proportion and the area abnormal parameter. The area radiation proportion = the number of abnormal points with a processing radiation coefficient greater than the preset processing radiation coefficient in the reference evaluation range of the abnormal point / the number of abnormal points in the reference evaluation range of the abnormal point.

[0088] In the method, the image is processed for a single target, if the target processing image completes the optimization analysis, i.e., the abnormal sub-regions or abnormal points complete the determination of the denoising execution index, the denoising processing is performed based on the denoising execution index of each abnormal sub-region or each abnormal point, for a single abnormal sub-region or a single abnormal point, if the denoising execution index is greater than a preset denoising execution index, the non-local mean algorithm is used for denoising processing, and the smoothing parameter used in the denoising processing process has a positive correlation with the denoising execution index, if the denoising execution index is less than or equal to the preset denoising execution index, the bilateral filtering algorithm is used for denoising processing, and the value range standard deviation used in the denoising processing process has a positive correlation with the denoising execution index.

[0089] The preset denoising execution index can be determined by the user according to the actual working scene, for example, the user can set it according to the image optimization record, the higher the requirement of the user for the effectiveness of the target processing image after the denoising processing, the greater the value of the preset denoising execution index, a method for determining the value of the preset denoising execution index is provided, the image optimization record in which the non-local mean algorithm is used for denoising processing is recorded as an execution reference record, and the minimum value of the denoising execution index in the execution reference record meeting the requirement of the user for the effectiveness of the target processing image after the denoising processing is recorded as the preset denoising execution index.

[0090] The technical scheme of the present application has been described in combination with the preferred embodiments shown in the drawings, but those skilled in the art can easily understand that the protection scope of the present application is obviously not limited to these specific embodiments. Those skilled in the art can make equivalent changes or replacements to the related technical features without departing from the principles of the present application, and the technical scheme after the changes or replacements will fall within the protection scope of the present application.

[0091] The above description is only the preferred embodiments of the present application and is not used to limit the present application; for those skilled in the art, the present application can have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. An intelligent noise reduction method for archival pictures, characterized in that, Comprise: According to the abnormal proportion index and the abnormal distribution index, the optimization analysis strategy of each target processing image is determined as auxiliary feature analysis or abnormal radiation analysis on the target processing image; When the auxiliary feature analysis is performed on the target processing image with the abnormal proportion index greater than the preset abnormal proportion index or the abnormal distribution index greater than the preset abnormal distribution index, the setting mode of the feature analysis set of each one-type target processing image and the auxiliary analysis mode are determined according to the correlation richness coefficient, the setting mode of the feature analysis set is determined based on the feature distribution correlation coefficient or the storage parameter correlation degree to determine the feature analysis set of the one-type target processing image, and the auxiliary analysis mode is determined based on the set reference key index or the defect confidence degree to determine the denoising processing mode to obtain the denoising execution index of each abnormal sub-region; When the abnormal radiation analysis is performed on the target processing image with the abnormal proportion index less than or equal to the preset abnormal proportion index and the abnormal distribution index less than or equal to the preset abnormal distribution index, the denoising execution mode of each abnormal point in each two-type target processing image is determined according to the processing radiation coefficient, which is based on the processing radiation coefficient and the area radiation coefficient, or the area radiation proportion and the area abnormal parameter to determine the denoising execution index of the abnormal point; Based on the determined denoising execution index, denoising processing is performed on each target processing image; For a single target processing image, the abnormal proportion index = the number of abnormal point positions contained in the target processing image / the number of point positions to be analyzed contained in the target processing image, the abnormal distribution index = the number of abnormal sub-regions existing in the target processing image / the number of sub-image regions existing in the target processing image, and the sub-image region with an abnormal point position is recorded as an abnormal sub-region; For a single one-type target processing image, the correlation richness coefficient is determined according to the page richness index and the one-type page proportion, and the correlation richness coefficient = ln (page richness index * one-type page proportion), the page richness index is the number of information pages existing in the target processing file to which the one-type target processing image belongs, and the one-type page proportion = the number of one-type target processing images corresponding to the information pages existing in the target processing file to which the one-type target processing image belongs / the number of information pages existing in the target processing file to which the one-type target processing image belongs; If the correlation richness coefficient of the one-type target processing image is greater than the preset correlation richness coefficient, the feature analysis set is determined based on the feature distribution correlation coefficient; If the correlation richness coefficient of the one-type target processing image is less than or equal to the preset correlation richness coefficient, the feature analysis set of the one-type target processing image is determined based on the storage parameter correlation degree.

2. The method of claim 1, wherein, The target processing image with the abnormal proportion index greater than the preset abnormal proportion index or the abnormal distribution index greater than the preset abnormal distribution index is recorded as a one-type target processing image.

3. The method of claim 1, wherein, The target processing image with the abnormal proportion index less than or equal to the preset abnormal proportion index and the abnormal distribution index less than or equal to the preset abnormal distribution index is recorded as a two-type target processing image.

4. The method of claim 2, wherein, The set reference key index of each abnormal sub-region is determined according to the abnormal reference index and the reference gradient index, and the noise reduction processing mode of each abnormal sub-region is determined based on the set reference key index.

5. The method of claim 4, wherein, If the set reference key index of an abnormal sub-region is greater than the preset set reference key index, the denoising execution index of the abnormal sub-region is determined according to the set reference key index and the set reference index; If the set reference key index of an abnormal sub-region is less than or equal to the preset set reference key index, the denoising execution index of the abnormal sub-region is determined according to the reference gradient index.

6. The method of claim 5, wherein, The defect confidence degree is determined according to the edge index difference value and the set verification index, and the noise reduction processing mode of each abnormal sub-region is determined based on the defect confidence degree.

7. The method of claim 6, wherein, If the defect confidence degree of a target processing image of a type is greater than the preset defect confidence degree, the denoising execution index of each abnormal sub-region is determined according to the defect edge parameter and the defect confidence degree; If the defect confidence degree of a target processing image of a type is less than or equal to the preset defect confidence degree, the denoising execution index of each abnormal sub-region is determined according to the reference gradient index and the defect confidence degree.

8. The method of claim 3, wherein, Under abnormal radiation analysis conditions, the processing radiation coefficient of each abnormal point in a target processing image of a type is determined according to the region gradient index and the gradient key index, and the noise reduction execution mode of each abnormal point is determined according to the processing radiation coefficient; The processing radiation coefficient is positively correlated with the region gradient index and the gradient key index, respectively. The abnormal radiation analysis condition is that the target processing image is optimized and analyzed by abnormal radiation analysis.

9. The method of claim 8, wherein, For a single target processing image of a type, if the processing radiation coefficient of an abnormal point is greater than the preset processing radiation coefficient, the denoising execution index of the abnormal point is determined based on the processing radiation coefficient and the region radiation coefficient; The denoising execution index is positively correlated with the processing radiation coefficient, and negatively correlated with the region radiation coefficient.

10. The method of claim 9, wherein, For a single target processing image of a type, if the processing radiation coefficient of an abnormal point is less than or equal to the preset processing radiation coefficient, the denoising execution index of the abnormal point is determined based on the region radiation proportion and the region abnormal parameter; The denoising execution index is negatively correlated with the region radiation proportion and the region abnormal parameter, respectively.

Images