Eye fundus image intelligent enhancement method and system

Abstract

The invention relates to the technical field of image data processing, in particular to an eye fundus image intelligent enhancement method and system. The method comprises the following steps: acquiring a grayscale image of a fundus image to be enhanced, and dividing the grayscale image into a plurality of image areas to determine an edge communication area formed by edge pixel points in the image areas; determining an edge complexity value of the image region; obtaining a gray complexity value of the image area; multiplying the edge complexity value of the target image area, the gray level complexity value and the frequency of the gray level in the target image area to obtain a first parameter value of the gray level corresponding to the target image area so as to obtain a second parameter value of the gray level, and multiplying the second parameter value of the gray level by the frequency of the gray level in the gray level image to obtain a second parameter value of the gray level; and the weighted frequency of the gray level is obtained so as to carry out histogram equalization on the gray level image to obtain an enhanced image. According to the technical scheme, the enhanced image which is more convenient for observing the fundus condition can be obtained.

Description

Technical Field

[0001] The present application relates to the technical field of image data processing, and in particular to a method and system for intelligently enhancing fundus images. Background Art

[0002] Fundus images refer to images obtained by capturing the fundus of a patient through image acquisition devices such as a fundus camera. Fundus images mainly include the retina, optic disc, retinal blood vessels, and macula of the patient's eye. Fundus images have a high reference value for screening patients for diseases such as diabetic retinopathy, glaucoma, hypertensive retinopathy, and macular degeneration.

[0003] When collecting fundus images, the contrast of the collected fundus images may be low due to insufficient light, limited shooting conditions, or reflections from the eyes themselves, making it inconvenient to observe the patient's eye condition. Therefore, it is necessary to perform image enhancement processing on the fundus images.

[0004] In order to enhance the fundus image, a Chinese patent application document with publication number CN112819730A provides an image enhancement processing method, including: S1: providing an image to be processed and obtaining a processing area of ​​the image to be processed; S2: obtaining a histogram of the processing area; S3: determining a minimum threshold and a maximum threshold of the grayscale value of the processing area; the minimum threshold is denoted as hmin, and the maximum threshold is denoted as hmax; S4: remapping the grayscale value of the processing area, when hmax>hmin, remapping the grayscale value between 0-1023 to between 0-255 to obtain a processed image.

[0005] When the related art performs enhancement processing on the fundus image, it is only performed based on the maximum grayscale threshold and the minimum grayscale threshold in the image, which makes it difficult to achieve effective enhancement of the fundus image. Summary of the invention

[0006] In order to overcome the problem that it is difficult to effectively enhance fundus images in the related art, the present application provides a fundus image intelligent enhancement method and system.

[0007] According to a first aspect of an embodiment of the present application, a method for intelligent enhancement of fundus images is provided, comprising: obtaining a grayscale image of a fundus image to be enhanced, and dividing the grayscale image into multiple image areas to determine an edge connected area composed of edge pixels in the image area; determining an edge complexity value of the image area according to the number of edge connected areas in the image area and the number of edge pixels in the same connected area; obtaining a grayscale complexity value of the image area, wherein the grayscale complexity value is used to characterize the complexity of the grayscale value in the image area; for a target image area in multiple image areas, multiplying the edge complexity value, the grayscale complexity value, and the frequency of the grayscale level in the target image area of ​​the target image area to obtain a first parameter value of the grayscale level corresponding to the target image area; summing all first parameter values ​​of the grayscale level corresponding to all target image areas to obtain a second parameter value of the grayscale level, and multiplying the second parameter value of the grayscale level by the frequency of the grayscale level in the grayscale image to obtain a weighted frequency of the grayscale level; performing histogram equalization on the grayscale image according to the weighted frequencies of all grayscale levels in the grayscale image, and using the image after the histogram equalization as an enhanced image.

[0008] In this way, the grayscale image is enhanced according to the weighted frequency of all gray levels in the grayscale image. Since the weighted frequency of the grayscale is determined according to the frequency of the grayscale in the target image area where it appears, the edge complexity value of the target image area and the grayscale complexity value, the edge complexity value can characterize the complexity of edge features such as blood vessels in the image area of ​​the fundus, and the grayscale complexity value can characterize the complexity of the grayscale values ​​in the image area, the weighted frequency of the grayscale can better reflect the characteristics of the grayscale, thereby obtaining an enhanced image that is more convenient for observing the fundus condition, and effectively realizing the enhancement of the fundus image.

[0009] Optionally, the edge complexity value of the image region is determined by: ,in, is the edge complexity value of the image area, B is the number of edge connected areas in the image area, is the indicator function with the natural constant as base, is the number of edge pixels in the i-th edge-connected region of the image region.

[0010] In this way, compared with other positions outside the edge line in the image area, the edge line in the image area has a higher probability of corresponding to the blood vessels of the eye. By determining the edge complexity value of the image area, it is helpful to characterize the blood vessel characteristics in the image area through the edge complexity value.

[0011] Optionally, the grayscale complexity value of the image region is determined by: , where H is the grayscale complexity value of the image area, D is the information entropy of the grayscale value of the image area, and N is the number of pixels in the image area. is the gray value of the mth pixel in the image area, is the average gray value of the gray values ​​of the pixels in the image area, is the standard deviation of the grayscale values ​​of pixels in the image area.

[0012] In this way, the grayscale complexity value of the image region can integrate the information entropy of the grayscale values ​​in the image region and the difference of the grayscale values ​​in the image region, so that the grayscale complexity value can better characterize the complexity of the grayscale values ​​in the image region.

[0013] Optionally, the information entropy of the grayscale value of the image region is determined by: , where D is the information entropy of the grayscale value of the image area, is the number of gray levels in the image area, is the frequency of occurrence of the g-th gray level in the image area, and log is the logarithmic function.

[0014] In this way, the information entropy of the grayscale value of the image region is determined according to the frequency of occurrence of different grayscale levels in the image region, and the complexity of the grayscale in the image region can be characterized by the information entropy.

[0015] Optionally, histogram equalization is performed on the grayscale image according to the weighted frequencies of all grayscale levels in the grayscale image, including: for a target grayscale level among all grayscale levels in the grayscale image, a cumulative distribution function value corresponding to the target grayscale level is determined according to the accumulated sum of the weighted frequencies from the minimum grayscale level among all grayscale levels to the target grayscale level; according to the cumulative distribution function value corresponding to the target grayscale level, the grayscale level after equalization of the target grayscale level is determined, and an image composed of pixel points corresponding to all the equalized grayscale levels is used as the image after histogram equalization of the grayscale image.

[0016] Optionally, determining the cumulative distribution function value corresponding to the target gray level according to the accumulated sum of weighted frequencies from the minimum gray level to the target gray level among all gray levels includes: , is the cumulative distribution function value corresponding to the target gray level. The target gray level ranks A from small to large among all the gray levels in the gray image. is the weighted frequency corresponding to the ath gray level in the grayscale image from small to large, and H is the number of grayscale types in the grayscale image.

[0017] Optionally, the grayscale after equalization of the target grayscale is determined according to the cumulative distribution function value corresponding to the target grayscale, including: obtaining the total number of grayscales to be mapped; multiplying the cumulative distribution function value corresponding to the target grayscale by the total number of grayscales to obtain the grayscale after equalization of the target grayscale.

[0018] In this way, since the greater the weighted frequency corresponding to the target gray level, the greater the difference in gray level between the target gray level and the gray level before the target gray level, and the greater the weighted frequency corresponding to the target gray level, the greater the probability that the pixel point with a gray value of the target gray level is key information, therefore, the difference in gray level between key information and non-key information in the grayscale image of the fundus can be increased, and the eye condition can be better observed through enhanced images.

[0019] According to a second aspect of an embodiment of the present application, a fundus image intelligent enhancement system is provided, comprising: a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the steps of the fundus image intelligent enhancement method provided in the first aspect of the present application are implemented.

[0020] The technical solution provided by the embodiments of the present application may include the following beneficial effects: enhancing the grayscale image according to the weighted frequency of all grayscale levels in the grayscale image; since the weighted frequency of the grayscale level is determined based on the frequency of the grayscale level in the target image area where it appears, the edge complexity value of the target image area, and the grayscale complexity value, the edge complexity value can characterize the complexity of edge features such as blood vessels in the image area of ​​the fundus, and the grayscale complexity value can characterize the complexity of the grayscale values ​​in the image area, the weighted frequency of the grayscale level can better reflect the characteristics of the grayscale level, thereby obtaining an enhanced image that is more convenient for observing the fundus condition, and effectively achieving the enhancement of the fundus image.

[0021] It should be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0023] Figure 1 is a flow chart of a method for intelligently enhancing fundus images according to an exemplary embodiment; Figure 2 The figure is a schematic diagram of the structure of a fundus image intelligent enhancement system according to an exemplary embodiment. DETAILED DESCRIPTION

[0024] Here, exemplary embodiments are described in detail, and examples thereof are shown in the accompanying drawings. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The implementations described in the following exemplary embodiments do not represent all implementations consistent with the present application.

[0025] First, the application scenario of the embodiment of the present application is briefly introduced. In the application scenario of the present application, in order to facilitate the observation of the patient's eye condition through fundus images, the collected fundus images can be enhanced. In related technologies, the fundus images are grayscale mapped according to the maximum grayscale threshold and the minimum grayscale threshold in the fundus images.

[0026] When performing grayscale mapping on fundus images, the related art does not consider the importance of grayscale in the grayscale image, so the obtained enhanced image may still not be convenient for observing the patient's eye condition, making it difficult to achieve effective enhancement of the fundus image.

[0027] In view of the above technical problems, the present application provides a method and system for intelligently enhancing fundus images. Figure 1 is a flow chart of a method for intelligently enhancing fundus images according to an exemplary embodiment. Figure 1 As shown, the method includes the following steps.

[0028] In step S101, a grayscale image of a fundus image to be enhanced is acquired, and the grayscale image is divided into a plurality of image regions to determine edge connected regions formed by edge pixels in the image regions.

[0029] The fundus image of the patient's eye can be collected by an image acquisition device such as a fundus camera, and the fundus image to be enhanced can be gray-scaled to obtain a grayscale image; alternatively, the fundus image to be enhanced can be selected from a database that pre-stores the fundus images of the patient and the selected fundus image to be enhanced can be obtained from the database, thereby gray-scaled to obtain a grayscale image.

[0030] The grayscale image can be divided into multiple image regions. The size of the image region can be 9×9, 11×11, 15×15, etc. Since the features in different image regions of the grayscale image are different, by dividing the grayscale image, it is easy to determine the features of different image regions in the grayscale image, so as to better realize the image enhancement processing of the grayscale image.

[0031] Edge detection can be performed on the image area to obtain edge pixel points in the image area. Since there are scattered blood vessels in the grayscale image of the fundus, these scattered blood vessels appear as edge lines in the grayscale image. In order to facilitate the observation of blood vessels in the patient's fundus, the edge connected area composed of edge pixel points in the image area can be determined.

[0032] The same edge connected region includes multiple connected edge pixels located in the image region, and the image operator for edge detection of the image region of the grayscale image can be an operator such as a Sobel operator, a Prewitt operator, a Canny operator, and a Laplacian operator.

[0033] In step S102, the edge complexity value of the image region is determined according to the number of edge connected regions in the image region and the number of edge pixels in the same connected region.

[0034] The more edge connected areas there are in the same image area, the more blood vessels there are in the fundus position corresponding to the image area; the more pixels there are in the same edge connected area, the longer or wider the blood vessels in the fundus position corresponding to the edge connected area.

[0035] The edge complexity value of the image area is determined based on the number of edge connected areas in the image area and the number of edge pixels in the same connected area. The edge complexity value can characterize the complexity of the edge lines in the image area. Since the edge lines in the image area correspond to the blood vessels of the eye, the edge complexity value of the image area can characterize the complexity of the blood vessels at the corresponding position of the image area in the fundus.

[0036] In one embodiment, the edge complexity value of the image region is determined in the following manner: ,in, is the edge complexity value of the image area, B is the number of edge connected areas in the image area, is the indicator function with the natural constant as base, is the number of edge pixels in the i-th edge-connected region of the image region.

[0037] In this way, compared with other positions outside the edge line in the image area, the edge line in the image area has a higher probability of corresponding to the blood vessels of the eye. By determining the edge complexity value of the image area, it is helpful to characterize the blood vessel characteristics in the image area through the edge complexity value.

[0038] In step S103, the grayscale complexity value of the image area is obtained.

[0039] The grayscale complexity value is used to characterize the complexity of the grayscale values ​​in the image area. For the image area in the grayscale image of the fundus, the larger the grayscale complexity value of the image area, the higher the complexity of the grayscale values ​​in the image area, and the more information about the eye is usually contained in the image area. The grayscale values ​​of the same part in the grayscale image of the fundus of the patient are relatively close. The grayscale levels that appear in the image area with higher complexity usually correspond to the key information of the fundus.

[0040] The key information in the grayscale image of the fundus may be the blood vessels, optic disc, and macula of the fundus; the non-key information in the grayscale image of the fundus may refer to the background part of the fundus.

[0041] On the contrary, the smaller the grayscale complexity value of the image area, the lower the complexity of the grayscale value in the image area, the less information about the eye is usually contained in the image area, and the grayscale values ​​of the same part in the grayscale image of the patient's fundus are relatively close. The grayscale levels that appear in the image area with lower complexity usually correspond to non-critical information such as the background part of the fundus.

[0042] In one embodiment, the grayscale complexity value of the image region is determined in the following manner: , where H is the grayscale complexity value of the image area, D is the information entropy of the grayscale value of the image area, and N is the number of pixels in the image area. is the gray value of the mth pixel in the image area, is the average gray value of the gray values ​​of the pixels in the image area, is the standard deviation of the grayscale values ​​of pixels in the image area.

[0043] By determining the absolute value of the difference between the grayscale value of the pixel in the image area and the average grayscale value of the entire image area, and dividing the absolute value of the difference by the standard deviation, the degree of difference between the grayscale values ​​of different pixels in the image area can be characterized; for the image area in the grayscale image of the fundus, the image area with higher grayscale complexity usually contains key information in the fundus; on the contrary, the image area with lower grayscale complexity usually contains non-key information in the fundus.

[0044] By dividing the absolute value of the difference between the grayscale value of the pixel and the average grayscale value by the standard deviation of the grayscale value of the image area, the grayscale difference is standardized, which can facilitate the comparison of grayscale complexity values ​​between different image areas, so that the obtained grayscale complexity value can better characterize the complexity of the grayscale values ​​in the image area.

[0045] The information entropy of the grayscale value of the image area can characterize the complexity of different grayscale levels in the image area. In the image area of ​​the grayscale image of the fundus, the image area with a higher grayscale complexity requires a higher degree of attention in order to observe the patient's eye condition.

[0046] In this way, the grayscale complexity value of the image region can integrate the information entropy of the grayscale values ​​in the image region and the difference of the grayscale values ​​in the image region, so that the grayscale complexity value can better characterize the complexity of the grayscale values ​​in the image region.

[0047] In one embodiment, the information entropy of the grayscale value of the image region is determined by: , where D is the information entropy of the grayscale value of the image area, is the number of gray levels in the image area, is the frequency of occurrence of the g-th gray level in the image area, and log is the logarithmic function.

[0048] In the grayscale image of the eye, multiple gray levels may appear in the same image area. As the number of gray levels in the image area increases, the complexity of the gray levels in the image area increases, and the position of the image area corresponding to the eye needs more attention.

[0049] Since the frequency of gray levels in the image area is between 0 and 1, is a negative number. The constraint of the negative sign can ensure that the value of information entropy is positive. When the number of pixels corresponding to the multiple gray levels appearing in the same image area is the same, When the value of is the largest and the number of pixels of all gray levels in the image area is the same, = , therefore, As the denominator of grayscale information entropy, it can realize the normalization of information entropy and ensure that the value of grayscale information entropy is between 0 and 1.

[0050] In this way, the information entropy of the grayscale value of the image region is determined according to the frequency of occurrence of different grayscale levels in the image region, and the complexity of the grayscale in the image region can be characterized by the information entropy.

[0051] In step S104, for a target image region among the multiple image regions, the edge complexity value, the grayscale complexity value and the frequency of the grayscale in the target image region are multiplied to obtain a first parameter value of the grayscale corresponding to the target image region.

[0052] Since the blood vessels in the grayscale image of the fundus mainly appear as scattered edge lines, the greater the edge complexity value of the target image area, the greater the probability that the target image area corresponds to the blood vessels of the fundus. Alternatively, the more complex the blood vessels at the position corresponding to the target image area in the fundus, the higher the probability that the pixel points corresponding to the grayscale level correspond to the blood vessels of the fundus.

[0053] On the contrary, the smaller the edge complexity value of the target image area is, the smaller the probability that the target image area corresponds to the blood vessels of the fundus is, or the blood vessels at the position corresponding to the target image area in the fundus are more simple. The pixel points corresponding to the gray level have a lower probability of corresponding to the blood vessels of the fundus.

[0054] The greater the grayscale complexity value of the target image area where the grayscale level in the grayscale image of the fundus is located, the higher the probability that the grayscale level corresponds to key information in the grayscale image of the fundus, for example, the grayscale level has a higher probability of corresponding to blood vessels, optic disc, and macula in the fundus; on the contrary, the smaller the grayscale complexity value of the target image area where the grayscale level in the grayscale image of the fundus is located, the lower the probability that the grayscale level corresponds to key information in the grayscale image of the fundus, for example, the grayscale level corresponds to the background area of ​​the fundus.

[0055] The product obtained by multiplying the edge complexity value, grayscale complexity value and frequency of grayscale in the target image area can better characterize the probability that the grayscale corresponds to key information in the fundus. Moreover, since the edge complexity value takes into account the shape feature of the edge line in the image area, and the background part of the fundus usually does not have the shape feature of the edge line, it helps to expand the contrast between the background part and other parts outside the background part in the subsequent enhanced image.

[0056] Since the first parameter value of the grayscale corresponding to the target image area also integrates the frequency of the grayscale appearing in the target image area, the larger the first parameter value of the grayscale corresponding to the target image area is, the higher the probability that the grayscale corresponds to more critical information in the target image area; and when the first parameter value of the grayscale corresponding to the target image area is larger, the target image area has a higher probability of being the key information in the grayscale image of the fundus, and when the grayscale has a higher probability of corresponding to more critical information in the target image area, the pixel point corresponding to the grayscale has a higher probability of corresponding to the key information in the grayscale image of the fundus.

[0057] In step S105, all first parameter values ​​corresponding to the grayscale of all target image regions are summed to obtain a second parameter value of the grayscale, and the second parameter value of the grayscale is multiplied by the frequency of the grayscale in the grayscale image to obtain a weighted frequency of the grayscale.

[0058] Since the first parameter value of the grayscale corresponding to the target image area can characterize the probability that the pixel point corresponding to the grayscale corresponds to the key information in the grayscale image of the fundus, by summing up all the first parameter values ​​of the grayscale corresponding to all target image areas, the first parameter values ​​of the same grayscale in all grayscale images of the grayscale image can be combined, so as to better characterize the probability that the pixel point corresponds to the key information in the grayscale image of the fundus through the second parameter value of the grayscale.

[0059] Multiplying the second parameter value of the grayscale by the frequency of the grayscale in the grayscale image can facilitate processing different pixels corresponding to the same grayscale in the grayscale image together, and when the grayscale image is subsequently histogram equalized, it is convenient to map the grayscale to a new grayscale.

[0060] In step S106, the grayscale image is histogram equalized according to the weighted frequencies of all gray levels in the grayscale image, and the image after the histogram equalization is used as the enhanced image.

[0061] When determining the weighted frequency of gray levels in a grayscale image, gray levels with higher weighted frequencies have a higher probability of corresponding to key information of the fundus, and gray levels with lower weighted frequencies have a lower probability of corresponding to key information of the fundus.

[0062] Compared with grayscale histogram based only on the frequency of occurrence of gray levels in the grayscale image, since the histogram equalization of the grayscale image is performed based on the weighted frequency of the grayscale, the weighted frequency of the grayscale is equal to the product of the second parameter value of the grayscale and the frequency of the grayscale in the grayscale image.

[0063] For two gray levels that appear with the same frequency in a grayscale image, when the second parameter values ​​of the two gray levels are different, the contrast of the two gray levels with different second parameter values ​​in the image after histogram equalization can be improved, so as to better distinguish the pixel points corresponding to the two gray levels with different second parameter values ​​in the enhanced image.

[0064] For example, since the weighted frequency of the grayscale corresponding to the background part of the fundus is smaller, it is possible to better distinguish between key information such as the blood vessels, optic disc and macula of the fundus and the background part in the enhanced image.

[0065] In one embodiment, a grayscale image is histogram equalized according to the weighted frequencies of all grayscale levels in the grayscale image, including: for a target grayscale level among all grayscale levels in the grayscale image, a cumulative distribution function value corresponding to the target grayscale level is determined according to the accumulated sum of the weighted frequencies from the minimum grayscale level among all grayscale levels to the target grayscale level; according to the cumulative distribution function value corresponding to the target grayscale level, a grayscale level after equalization of the target grayscale level is determined, and an image composed of pixel points corresponding to all grayscale levels after equalization is used as an image after histogram equalization of the grayscale image.

[0066] For example, the grayscale level in a grayscale image can range from 0 to 255, and there are 256 grayscale levels in total; the grayscale level in the collected grayscale image of the fundus may range from 0 to 120, and the difference between the maximum grayscale level and the minimum grayscale level in the grayscale image of the fundus is 120. After histogram equalization processing of the grayscale image of the fundus, the difference between the maximum grayscale level and the minimum grayscale level in the grayscale image of the fundus can be increased from the original 120 to 256.

[0067] For another example, when the grayscale range of a grayscale image is between 0 and 120, if the weighted frequencies of different grayscales in the grayscale image are the same, the grayscale difference between a pixel with a grayscale of 64 and a pixel with a grayscale of 8 is 56; when the grayscale image is histogram equalized to a grayscale range of 0 to 255, the grayscale of a pixel with a grayscale of 64 in the original image in the enhanced image is (255 / 120)×64=136, and the grayscale of a pixel with a grayscale of 8 in the original image in the enhanced image is (255 / 120)×8=17. The grayscale difference between the two grayscale pixels is enlarged to 136-17=119, and the two grayscale pixels can be more easily distinguished in the enhanced image.

[0068] Through the intelligent fundus image enhancement method provided in the embodiment of the present application, the grayscale image is enhanced according to the weighted frequency of all grayscale levels in the grayscale image. Since the weighted frequency of the grayscale level is determined based on the frequency of the grayscale level in the target image area where it appears, the edge complexity value of the target image area and the grayscale complexity value, the edge complexity value can characterize the complexity of edge features such as blood vessels in the image area of ​​the fundus, and the grayscale complexity value can characterize the complexity of the grayscale values ​​in the image area, the weighted frequency of the grayscale level can better reflect the characteristics of the grayscale level, thereby obtaining an enhanced image that is more convenient for observing the fundus condition, and effectively realizing the enhancement of the fundus image.

[0069] In one embodiment, determining the cumulative distribution function value corresponding to the target gray level according to the accumulated sum of the weighted frequencies from the minimum gray level to the target gray level among all gray levels includes: , is the cumulative distribution function value corresponding to the target gray level. The target gray level ranks A from small to large among all the gray levels in the gray image. is the weighted frequency corresponding to the ath gray level in the grayscale image from small to large, and H is the number of grayscale types in the grayscale image.

[0070] Here, since the cumulative distribution function value corresponding to the target gray level is determined based on the weighted frequency corresponding to the target gray level, and the weighted frequency corresponding to the target gray level is obtained based on the edge complexity value, gray complexity value and frequency of the target gray level in the target image area, the weighted frequency corresponding to the target gray level can characterize the probability of key information of pixels with a gray value of the target gray level in the gray image of the fundus.

[0071] For example, when the number of gray levels in the grayscale image of the fundus before image enhancement processing is 5, the cumulative distribution function value corresponding to the smallest gray level is S1=M1 / (M1+M2+M3+M4+M5), the cumulative distribution function value corresponding to the second to last gray level of all gray levels is S2=(M1+M2) / (M1+M2+M3+M4+M5), and the cumulative distribution function value corresponding to the third to last gray level of all gray levels is S3=(M1+M2+M3) / (M1+M2+M3+M4+M5).

[0072] It can be seen that the greater the weighted frequency corresponding to the target gray level, the greater the difference between the cumulative distribution function value of the target gray level and the cumulative distribution function value of the previous gray level of the target gray level. Since the cumulative distribution function value of the gray level can determine the gray level after mapping in the enhanced image, the greater the weighted frequency corresponding to the target gray level, the greater the difference between the gray level of the target gray level after mapping and the gray level of the previous gray level of the target gray level after mapping, which is more conducive to the human eye or image recognition equipment to identify the pixel points with the gray value of the target gray level from the enhanced image.

[0073] And because the gray level with a larger weighted frequency has a greater probability of being the key information in the grayscale image of the fundus, therefore, the larger the weighted frequency corresponding to the target gray level, the more conducive it is for the human eye or image recognition equipment to identify the key information in the fundus from the enhanced image, thereby making judgments on the eye information.

[0074] In one embodiment, the grayscale level after equalization of the target grayscale level is determined according to the cumulative distribution function value corresponding to the target grayscale level, including: obtaining the total number of grayscale levels to be mapped; multiplying the cumulative distribution function value corresponding to the target grayscale level by the total number of grayscale levels to obtain the grayscale level after equalization of the target grayscale level.

[0075] The total number of gray levels to be mapped is obtained and can be set according to actual needs. For example, the total number of gray levels to be mapped can be set to 256 to cover every non-negative gray level between 0 and 255.

[0076] Since the cumulative distribution function value is determined based on the weighted frequency corresponding to the target gray level, the larger the weighted frequency corresponding to the target gray level, the greater the difference in the cumulative distribution function value of the target gray level relative to the previous gray level. Therefore, the larger the weighted frequency corresponding to the target gray level, the greater the difference in gray level between the target gray level and the gray level before the target gray level.

[0077] In this way, since the greater the weighted frequency corresponding to the target gray level, the greater the difference in gray level between the target gray level and the gray level before the target gray level, and the greater the weighted frequency corresponding to the target gray level, the greater the probability that the pixel point with a gray value of the target gray level is key information, therefore, the difference in gray level between key information and non-key information in the grayscale image of the fundus can be increased, and the eye condition can be better observed through enhanced images.

[0078] Figure 2 FIG. 1 is a schematic diagram showing a structure of a fundus image intelligent enhancement system 1000 according to an exemplary embodiment. Figure 2 The fundus image intelligent enhancement system 1000 includes: a processor 1100 and a memory 1200, wherein the memory 1200 stores computer program instructions, and when the computer program instructions are executed by the processor 1100, all or part of the steps of the fundus image intelligent enhancement method in the present application are implemented.

[0079] It should be understood that the features of some embodiments of the various applications described herein may be combined with each other, unless specifically stated otherwise.

[0080] Although terms such as "first", "second", and "third" may be used herein to describe various components, parts, regions, layers, or sections, these components, parts, regions, layers, or sections are not limited to these terms. Instead, these terms are only used to distinguish one component, part, region, layer, or section from another component, part, region, layer, or section. Therefore, without departing from the teachings of the examples described herein, the first component, part, region, layer, or section mentioned in the examples may also be referred to as the second component, part, region, layer, or section.

[0081] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In the description herein, the meaning of "plurality" is at least two, such as two, three, etc., unless otherwise clearly and specifically defined.

[0082] In addition, the word "exemplary" is used herein to mean serving as an example, instance, or diagram. Any aspect or design described herein as "exemplary" is not necessarily to be understood as being advantageous over other aspects or designs. Instead, the use of the word exemplary is intended to present concepts in a concrete manner. As used herein, the term "or" is intended to mean an inclusive "or" rather than an exclusive "or".

[0083] Likewise, although the present application has been shown and described with respect to one or more implementations, equivalent variations and modifications will occur to those skilled in the art upon reading and understanding this specification and the accompanying drawings. With particular regard to the various functions performed by the components (e.g., elements, resources, etc.) described above, unless otherwise noted, the terms used to describe such components are intended to correspond to any component (functionally equivalent) that performs the specific functions of the described components, even if not structurally equivalent to the disclosed structures.

[0084] Additionally, while particular features of the present application may have been disclosed with respect to only one of several implementations, such features may be combined with one or more other features of other implementations as may be desired and advantageous for any given or particular application.

[0085] Those skilled in the art will readily appreciate other embodiments of the present application after considering the specification and practicing the invention disclosed herein. The present application is intended to cover any modification, use or adaptation of the present application, which follows the general principles of the present application and includes common knowledge or customary technical means in the art that are not disclosed in the present application, and the specification and embodiments are only considered as exemplary.

[0086] It should be understood that the present application is not limited to the exact construction that has been described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof.

Claims

1. A fundus image intelligent enhancement method, characterized in that: include: Acquire a grayscale image of the fundus image to be enhanced, and divide the grayscale image into a plurality of image regions to determine an edge connected region composed of edge pixels in the image region; The edge complexity value of the image area is determined based on the number of edge connected areas in the image area and the number of edge pixels in the same connected area. , B is the number of edge connected regions in the image region, is the indicator function with the natural constant as base, is the number of edge pixels in the i-th edge-connected region of the image area; Get the grayscale complexity value of the image area , D is the information entropy of the gray value of the image area, N is the number of pixels in the image area, is the gray value of the mth pixel in the image area, is the average gray value of the gray values ​​of the pixels in the image area, is the standard deviation of the grayscale values ​​of pixels in the image area; For a target image region among the multiple image regions, multiply an edge complexity value, a grayscale complexity value, and a frequency of grayscale in the target image region to obtain a first parameter value of the grayscale corresponding to the target image region; Sum all first parameter values ​​of all target image regions corresponding to the grayscale to obtain a second parameter value of the grayscale, and multiply the second parameter value of the grayscale by the frequency of the grayscale in the grayscale image to obtain a weighted frequency of the grayscale; According to the weighted frequencies of all gray levels in the gray image, the gray image is histogram equalized, and the image after histogram equalization is used as the enhanced image.

2. The method for intelligent enhancement of fundus images according to claim 1, characterized in that: The information entropy of the grayscale values ​​of an image region is determined in the following way: , where D is the information entropy of the grayscale value of the image area, is the number of gray levels in the image area, is the frequency of occurrence of the g-th gray level in the image area, and log is the logarithmic function.

3. The method for intelligent enhancement of fundus images according to claim 1, characterized in that: According to the weighted frequency of all gray levels in the grayscale image, the grayscale image is histogram equalized, including: For a target gray level among all gray levels in the gray image, a cumulative distribution function value corresponding to the target gray level is determined according to the cumulative sum of weighted frequencies from the minimum gray level among all gray levels to the target gray level; According to the cumulative distribution function value corresponding to the target gray level, the gray level after equalization of the target gray level is determined, and an image composed of pixel points corresponding to all the equalized gray levels is used as an image after histogram equalization of the gray image.

4. The method for intelligent enhancement of fundus images according to claim 3, characterized in that: According to the cumulative sum of the weighted frequencies from the minimum gray level to the target gray level among all gray levels, the cumulative distribution function value corresponding to the target gray level is determined, including: , is the cumulative distribution function value corresponding to the target gray level. The target gray level ranks A from small to large among all the gray levels in the gray image. is the weighted frequency corresponding to the ath gray level in the grayscale image from small to large, and H is the number of grayscale types in the grayscale image.

5. The method for intelligent enhancement of fundus images according to claim 3, characterized in that: According to the cumulative distribution function value corresponding to the target gray level, the gray level after equalization of the target gray level is determined, including: Get the total number of gray levels that need to be mapped; The cumulative distribution function value corresponding to the target gray level is multiplied by the total number of gray levels to obtain a gray level after the target gray level is equalized.

6. A fundus image intelligent enhancement system, characterized in that: include: A processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the method for intelligent enhancement of fundus images according to any one of claims 1 to 5 is implemented.

Images