Fluorescence in situ hybridization image edge enhancement method based on illumination fusion and local binary

Abstract

The invention discloses a method for enhancing the contour of a fluorescence in situ hybridization image by illumination fusion and local binary, which relates to the field of image processing and solves the problems that the existing tissue cell fluorescence image preprocessing has image defects such as uneven local illumination, large noise, and different dye concentrations in the collected image. The method comprises: removing noise in the image by using a median filtering method, calculating a linear equation to obtain the texture weight of the image, analyzing and processing the texture weight and a smoothing function to obtain a brightness image and a smoothed image, and performing grayscale processing to obtain grayscale distribution information; selecting coefficients by using a local binary algorithm, enhancing the image, and obtaining an illumination enhanced image; performing weighted addition of the contour feature image and the illumination enhanced image, analyzing 8 neighborhoods around 8 maximum pixel values, identifying the area with high energy gradient by using a piecewise linear function and a heat map, and the high-frequency signal area is the contour area of ​​the fluorescence in situ hybridization tissue cell image.

Description

Technical Field

[0001] The invention relates to the technical field of image processing, and in particular to a fluorescence in situ hybridization image contour enhancement method based on illumination fusion and local binary value. Background Art

[0002] Fluorescence in situ hybridization (FISH) is an important molecular detection technology that uses fluorescently labeled nucleic acid probes to hybridize with specific target genes, so that the target genes present specific fluorescent signals under a fluorescence microscope. Tissue cells are a common sample type in fluorescence in situ hybridization technology. In order to accurately locate tissue cells, ultraviolet light is usually used to stain the cell nucleus in the hybridization experiment to obtain fluorescent images of tissue cells for subsequent analysis. However, when preprocessing the fluorescent images of tissue cells, it is still very difficult to extract the contour details of the cell image due to problems such as poor image quality, complex dye structure, uneven illumination, and different concentrations. The contour details are the key to the subsequent segmentation and positioning of tissue cells.

[0003] In recent years, in order to optimize and enhance the contour feature information of complex tissue images, many excellent contour enhancement algorithms and networks have been applied to the processing of tissue cell images. However, the problem of uneven illumination is still prominent. Although LBP and its improved algorithms have improved detail extraction, these methods are insufficient in optimizing the separation of uneven illumination and reflection, resulting in limited overall image quality and detail enhancement effects. Secondly, LBP and its improved methods show limitations in complex backgrounds and high-noise scenes, are easily interfered by noise, and the extracted detail features are not comprehensive enough, especially in the extraction of complex cell contours. It is difficult to achieve ideal results. Furthermore, although the multi-feature fusion algorithm improves the expression ability of contour information, it has high computational complexity and strong parameter sensitivity, and still shows insufficient robustness in high-noise images.

[0004] In addition, existing methods are not very adaptable to the characteristics of FISH tissue cells, the detail enhancement of bright and dark areas is easily unbalanced, the contour coherence is poor, and the noise suppression ability is also limited. Finally, the insufficient optimization of feature weighting schemes and poor adaptability to dynamic and complex scenes further limit the application scope of these methods.

[0005] Therefore, a new method is needed to comprehensively improve the above problems in order to enhance the contour features of FISH tissue cell images more efficiently and robustly. Summary of the invention

[0006] The present invention aims to solve the problems in the prior art of preprocessing tissue cell fluorescence images, such as uneven local illumination, large noise, different dye concentrations, etc. in the collected images. To solve the above technical problems, the present invention is implemented by the following technical solutions:

[0007] Solution 1: The present invention proposes a method for enhancing the contour of a fluorescence in situ hybridization image based on illumination fusion and local binary, the method comprising the following steps:

[0008] Step 1: Use the median filtering method to remove salt and pepper noise in the image, obtain the texture weight of the image by calculating the linear equation, use the texture weight and smoothing function to analyze and process the illumination and texture features in the image to obtain a brightness image and a smoothed image, and perform grayscale processing on the brightness image and the smoothed image to obtain grayscale distribution information;

[0009] Step 2: Based on the grayscale distribution information of the image, a local binary algorithm is used to select coefficients for extracting the contour feature image, and the image is enhanced to obtain a light-enhanced image;

[0010] Step 3: Perform weighted addition of the contour feature image and the illumination enhancement image, and use a thermal map to obtain an image with a high energy gradient;

[0011] Step 4: Analyze the eight neighborhoods around the eight maximum pixel values, set breakpoints for the linear function based on the analysis results, determine the optimization range of the linear function for the image, and identify areas with high energy gradients through piecewise linear functions and heat maps. The high-frequency signal area is the contour area of ​​the fluorescence in situ hybridization tissue cell image.

[0012] Furthermore, a preferred implementation is provided, in which the method for calculating the texture weight of the image by means of a linear equation in step 1 is: obtaining the texture weight by solving the constructed sparse matrix.

[0013] Furthermore, a preferred embodiment is provided, wherein the texture weight of the image in step 1 also includes the step of calculating the differential image texture weight in the vertical direction and the differential image texture weight in the horizontal direction.

[0014] Further, a preferred embodiment is provided, wherein the coefficient in step 2 is achieved by selecting the 25% and 75% quantiles.

[0015] Furthermore, a preferred implementation is provided, which calculates the vertical difference of the image, that is, the difference between each pixel and the next pixel, extends the vertical difference to the image boundary to form a periodic difference, calculates the horizontal difference of the image, processes the boundary pixels, and applies one-dimensional smoothing filtering to the horizontal and vertical differences respectively.

[0016] Furthermore, a preferred embodiment is provided, wherein the method for forming the periodic difference is: extending the difference in the vertical direction to the image boundary and connecting them in sequence from beginning to end to form a periodic difference.

[0017] Furthermore, a preferred implementation is provided, wherein the image enhancement processing in step 3 includes the steps of illumination smoothing, texture enhancement and brightness adjustment.

[0018] Furthermore, a preferred implementation is provided, wherein step 4 also includes the step of performing average filtering calculation on 16 neighborhoods of the image.

[0019] Solution 3: A computer device includes a processor and a storage medium, wherein the memory stores a computer program. When the processor runs the computer program stored in the memory, the processor executes any one of the methods described in Solution 1.

[0020] Solution 4: A computer storage medium for storing a computer program. When the computer program is read by a computer, the computer executes any one of the methods described in Solution 1.

[0021] The present invention is beneficial in that:

[0022] The method for enhancing the contour of fluorescence in situ hybridization images based on illumination fusion and local binary values ​​described in the present invention effectively solves the main problems in FISH tissue cell image enhancement by combining median filtering, illumination smoothing, improved local binary pattern LBP algorithm and thermal map energy gradient analysis. Median filtering retains the key features of the image while removing salt and pepper noise; the illumination smoothing image generated by texture weights and smoothing functions significantly improves the problem of uneven illumination; the dynamic adjustment of the LBP coefficient enhances the expression of image edges and details and reduces the discontinuity caused by extreme values; the clarity of contour features is further improved by weighted fusion of feature maps and illumination enhanced images; the high energy gradient area is identified by combining thermal maps and piecewise linear functions, effectively removing low-frequency noise and background impurities.

[0023] The method for contour enhancement of fluorescence in situ hybridization images based on illumination fusion and improved local binary patterns described in the present invention removes high-frequency noise in the image, such as random noise, through neighborhood mean filtering, retains the main structure and contour features of the cell, makes the image smoother and more natural, and provides a cleaner basis for subsequent analysis. Subsequently, the most prominent areas in the image, such as cell boundaries or high-contrast areas, are extracted, and local characteristics of their neighborhood statistics, such as minimum and maximum values, are calculated to achieve the ability to dynamically adjust processing parameters. Finally, the overall contrast and local details of the image are enhanced in combination with piecewise linear transformation, and the cell contours and structures are further highlighted by adapting to cell images with different brightness and contrast characteristics, providing high-quality visual input for analysis tasks such as segmentation of tissue cell images, target detection, and feature extraction.

[0024] The present invention is also applicable to the fields of medical diagnosis and tumor marker detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 Schematic diagram of original images of different fields of view after median filtering according to the first embodiment.

[0026] Figure 2 This is a schematic diagram of illumination enhancement of images with different fields of view as described in the first embodiment.

[0027] Figure 3 Schematic diagram of the local binary algorithm for different fields of view described in Implementation Method 1.

[0028] Figure 4 It is a schematic diagram of weighted addition of contour feature images of different fields of view and illumination enhancement images described in Implementation Method 1.

[0029] Figure 5 This is a thermal schematic diagram of different fields of view described in the first embodiment.

[0030] Figure 6 Schematic diagram of the final enhanced image of different fields of view described in the eleventh embodiment.

[0031] Figure 7 Schematic diagram of original images of different fields of view described in Implementation Example 11. DETAILED DESCRIPTION

[0032] In order to make the purpose, technical solutions and advantages of the implementation methods of the present application clearer, the technical solutions in the implementation methods of the present application will be clearly and completely described below in conjunction with the drawings in the implementation methods of the present application. Obviously, the described implementation methods are only part of the implementation methods of the present application, not all of the implementation methods.

[0033] Embodiment 1: This embodiment provides a method for enhancing the contour of a fluorescence in situ hybridization image based on illumination fusion and local binary value, and the method comprises the following steps:

[0034] Step 1: Use the median filtering method to remove salt and pepper noise in the image, obtain the texture weight of the image by calculating the linear equation, use the texture weight and smoothing function to analyze and process the illumination and texture features in the image to obtain a brightness image and a smoothed image, and perform grayscale processing on the brightness image and the smoothed image to obtain grayscale distribution information;

[0035] Step 2: Based on the grayscale distribution information of the image, a local binary algorithm is used to select coefficients for extracting the contour feature image, and the image is enhanced to obtain a light-enhanced image;

[0036] Step 3: Perform weighted addition of the contour feature image and the illumination enhancement image, and use a thermal map to obtain an image with a high energy gradient;

[0037] Step 4: Analyze the eight neighborhoods around the eight maximum pixel values, set breakpoints for the linear function based on the analysis results, determine the optimization range of the linear function for the image, and identify areas with high energy gradients through piecewise linear functions and heat maps. The high-frequency signal area is the contour area of ​​the fluorescence in situ hybridization tissue cell image.

[0038] Embodiment 2: This embodiment further limits the method for enhancing the contour of fluorescence in situ hybridization images with illumination fusion and local binary values ​​described in embodiment 1. The linear equation calculated in step 1 is solved by constructing a sparse matrix.

[0039] Implementation method three: This implementation method further limits the method for contour enhancement of fluorescence in situ hybridization images by illumination fusion and local binary value described in implementation method one. The texture weight of the image in step 1 also includes the steps of calculating the image texture weights of the vertical difference and the horizontal difference.

[0040] Embodiment 4: This embodiment further limits the method for enhancing the contour of fluorescence in situ hybridization images by combining illumination fusion and local binary values ​​described in embodiment 1. The coefficients in step 2 are achieved by selecting the 25% and 75% quantiles.

[0041] Implementation method five. This implementation method further limits the method for contour enhancement of fluorescence in situ hybridization images of illumination fusion and local binary value described in implementation method three. The vertical difference of the image, that is, the difference between each pixel and the next pixel, is calculated. The vertical difference is extended to the image boundary to form a periodic difference. The horizontal difference of the image is calculated, and the boundary pixels are processed. One-dimensional smoothing filtering is applied to the horizontal and vertical differences respectively.

[0042] Implementation method six: This implementation method further limits the illumination fusion and local binary fluorescence in situ hybridization image contour enhancement method described in implementation method five. The method for forming the periodic difference is: extending the vertical difference to the image boundary and connecting them in sequence from beginning to end to form a periodic difference.

[0043] Embodiment 7: This embodiment further limits the method for contour enhancement of fluorescence in situ hybridization images with illumination fusion and local binary value described in embodiment 1. The image enhancement processing in step 3 includes the steps of illumination smoothing, texture enhancement and brightness adjustment.

[0044] Embodiment 8: This embodiment further limits the method for enhancing the contour of fluorescence in situ hybridization images by illumination fusion and local binary value described in embodiment 1. Step 4 also includes a step of performing average filtering calculation on 16 neighborhoods of the image.

[0045] Embodiment 9. This embodiment proposes a computer device, including a processor and a storage medium, wherein a computer program is stored in the memory. When the processor runs the computer program stored in the memory, the processor executes the method described in any one of embodiments 1 to 8.

[0046] Embodiment 10: This embodiment proposes a computer storage medium for storing a computer program. When the computer program is read by a computer, the computer executes the method described in any one of Embodiments 1 to 8.

[0047] Implementation 11. This implementation includes an example, which is used to explain the above implementation. Figures 1 to 7 The present embodiment is described. The method for enhancing the contour of a fluorescence in situ hybridization image with illumination fusion and local binary value described in the present embodiment specifically comprises the following steps:

[0048] First, median filtering is used to remove most of the salt and pepper noise in the image, and filtering is performed to retain the characteristics of the image.

[0049] The texture weight of the image is obtained through a linear equation, and the smoothed image of the brightness map is obtained using the texture weight and a smoothing function.

[0050] The grayscale distribution information of the image is used to select coefficients for the LBP algorithm. This method can highlight the important features in the image, make the edges and details more prominent, and reduce the discontinuity caused by extreme values.

[0051] The contour feature image extracted by LBP is weightedly added to the illumination enhanced image to further enhance the contour features of the image. In order to further remove the influence of impurity information, the thermal map is used to obtain the image part with larger energy gradient.

[0052] The eight neighborhoods around the eight maximum pixel values ​​were analyzed, and breakpoints were set for the linear function based on the analysis results to determine the optimization range of the function for the image. Finally, the area with high energy gradient was identified through piecewise linear function and heat map, and the area with low energy gradient was effectively filtered out. The high-frequency signal area was the contour area of ​​the FISH tissue cell image.

[0053] In summary, the main problems in FISH tissue cell image enhancement were effectively solved by combining median filtering, light smoothing, improved local binary pattern (LBP) algorithm and heat map energy gradient analysis. Median filtering retains the key features of the image while removing salt and pepper noise; the light smoothing image generated by texture weight and smoothing function significantly improves the problem of uneven lighting; dynamic adjustment of LBP coefficient enhances the expression of image edges and details and reduces the discontinuity caused by extreme values; weighted fusion of feature map and light enhanced image further improves the clarity of contour features; combined with heat map and piecewise linear function to identify high energy gradient areas, low-frequency noise and background impurities are effectively removed.

[0054] By analyzing and processing the lighting and texture features in the image, dynamic range expansion and detail enhancement are achieved;

[0055] Analyze and process the illumination and texture features in the image according to the above;

[0056] Step 1. Import the required library files. NumPy: used for efficient array operations. Images are usually represented as two-dimensional or multi-dimensional arrays. NumPy provides operations on these arrays. Time: used to record the running time of the program to help analyze the performance of the code. Matplotlib.pyplot: used to draw images to help display the results of image processing. Cv2: OpenCV library, used for image processing, including operations such as reading images, converting colors, and image thresholding.

[0057] Step 2: Calculate the texture weight (vertical and horizontal) of the image to measure the texture difference between different pixels and use it for subsequent lighting smoothing. fin (image matrix), sigma (smoothing kernel size), sharpness (smoothing coefficient), calculate the texture weight of the image, which indicates the significance of the pixel value change, calculate the vertical difference of the image (the difference between each pixel and the next pixel), extend the vertical difference to the image boundary (connect end to end to form a periodic difference), calculate the horizontal difference of the image, and process the boundary pixels (connect end to end), apply one-dimensional smoothing filtering (using a mean kernel) to the horizontal and vertical differences respectively, reduce the influence of noise, and calculate the texture weight in the horizontal and vertical directions according to the formula: the weight value is inversely proportional to the difference and filtering result, and avoids division by zero (plus sharpness).

[0058] Calculate the texture weight of the image in the vertical and horizontal directions to measure the degree of change between pixels in the image. The texture weight value is used in subsequent lighting smoothing, which helps to preserve important image details and eliminate uneven lighting.

[0059] Step 3: For the two-dimensional array tmp, first transpose the matrix (tmp.T) and then flatten it into a one-dimensional column vector.

[0060] Step 4: Smooth the image illumination distribution by constructing and solving a sparse matrix linear equation group. In the illumination smoothing process, local details are retained and the problem of uneven brightness is eliminated. The optimized illumination components are generated to provide a basis for subsequent enhancement.

[0061] Calculate the illumination map based on the illumination weights to solve the image illumination smoothing problem. IN (input image), wx (horizontal texture weight), wy (vertical texture weight), lambd (smoothing parameter), solve the sparse matrix linear equation, smooth the illumination map, obtain the image size (number of rows r and number of columns c), and calculate the total number of pixels k, and calculate the horizontal and vertical differential terms. Calculate the differential terms under periodic boundary conditions, convert the periodic differentials into column vectors for constructing sparse matrices. Construct a sparse matrix to represent the horizontal and vertical differential terms, calculate the diagonal elements in the sparse matrix, construct the final linear system matrix A, solve the equations with the sparse matrix, obtain the output column vector, and convert the column vector back to image format as output.

[0062] Convert the two-dimensional image data into a one-dimensional column vector for matrix solution of linear equations. This step provides the basis for constructing sparse matrices and solving linear equations.

[0063] Step 5: Perform illumination smoothing on the image. I (image), lambd (smoothing parameter), sigma (smoothing kernel size), sharpness (sharpness coefficient). By calculating the texture weights wx and wy, call the function in step 4 and return the smoothed image S.

[0064] The texture weight and linear equation solution are used to smooth the illumination distribution of the input image, and the image after illumination smoothing is returned to provide smooth illumination information for subsequent image enhancement.

[0065] Step 6: Enhance the input image, including illumination smoothing, texture enhancement, and brightness adjustment, img_path (image path), mu (enhancement intensity), a, b (brightness adjustment parameters), read the image and normalize it, extract the maximum brightness channel of the image as the illumination component t_b, call the illumination smoothing function tsmooth, calculate the illumination map t, and generate the enhancement weight map W according to the parameter mu, enhance the original image, generate the enhanced image I2, and mark the area with lower brightness, call the maximum entropy enhancement algorithm to enhance the low brightness area, and combine it with the illumination weight to generate the enhancement result, add the enhancement results to get the final enhanced image. That is, the illumination enhancement result and the detail enhancement result are fused to generate the final optimized image.

[0066] According to the grayscale distribution information of the image, the coefficients are selected by using the LBP algorithm, and the contour feature image extracted by the LBP is weightedly added to the illumination enhanced image to further enhance the contour features of the image.

[0067] Step 1. Import the required library files, cv2 for image processing operations, NumPy library (np) for array operations and numerical calculations, and Matplotlib (pyplot) for data visualization.

[0068] Step 2: Input image img and center pixel value center, check whether the value of pixel point (x, y) is greater than or equal to the center pixel value, return 1 if greater than or equal, 0 if less than, use try-except to avoid array out of bounds. Used for local binary pattern (LBP) feature calculation to binarize pixels.

[0069] Step 3: Input the grayscale image img and pixel coordinates (x, y), calculate the LBP value of the pixel, compare the pixel values ​​in the upper left and lower right directions (left and right), and return the dynamic weight alpha or beta according to the comparison result. This is to achieve improved LBP calculation and enhance feature expression through dynamic weights.

[0070] Step 4: Count the distribution frequency of grayscale values ​​(0-255), convert the two-dimensional histogram into a one-dimensional array, calculate the cumulative distribution and normalize it to make it range [0, 1], and select the 25% and 75% quantiles as dynamic weight coefficients. That is, dynamically determine the weights alpha and beta to make the algorithm adaptive to images with different grayscale distributions.

[0071] Step 5: Receive a list output_list, which contains the image to be displayed and meta information. According to the type (image type) in each dictionary, display the grayscale image or histogram, dynamically calculate and return the weight coefficients alpha and beta, and provide parameters for subsequent LBP feature calculation. It is used to visualize the algorithm output (such as grayscale image, binary image, histogram, etc.).

[0072] Step 6. Define the image number num to facilitate file processing and storage. Use the imread function of OpenCV to read the two images. The path format is f-string. The parameter cv2.IMREAD_GRAYSCALE indicates that the image is loaded in grayscale. Check whether the two images are loaded successfully. If the loading fails, print an error message and exit the program. Check whether the dimensions (height, width) of the two images are consistent. If not, print an error message and exit the program.

[0073] Step 7. Define weighting parameters: alpha: weight of the first image (relatively large, dominates the characteristics of the output image), beta: weight of the second image (smaller, secondary influence), gamma: used to adjust the overall brightness of the output image (usually set to 0 for no adjustment). Perform weighted addition of the two images, the formula is as follows:

[0074] weighted_image(x,y)=image1(x,y)⋅α+image2(x,y)⋅β+γ

[0075] Combining the features of the two images, highlighting the characteristics of the primary image (image1) according to the weight ratio, while introducing the details of the secondary image (image2), α = 0.85, β = 0.15, γ = 0

[0076] According to the method described above, the area with high energy gradient is extracted through the heat map, and the 8 neighborhoods of the 8 maximum pixel values ​​are analyzed, and the linear function breakpoints are set based on the results.

[0077] Step 1. Import the required library files, numpy: for array operations and numerical calculations, matplotlib.pyplot and matplotlib.image: for drawing and saving images, scipy.ndimage.generic_filter: for applying custom filters to images (such as calculating neighborhood averages).

[0078] Step 2: Set the image number num to facilitate dynamic loading and processing of images with specific numbers. Use the matplotlib.image.imread function to load the image, check the image dimensions, and if the image is a color image (3 channels), calculate the average value of each channel and convert it to a grayscale image.

[0079] Step 3: Extract the 8 pixels with the highest grayscale values ​​in the image and their locations to prepare for subsequent analysis and visualization.

[0080] Flatten the image into a one-dimensional array, find all unique pixel values, and automatically sort them in ascending order, get the maximum 8 pixel values, and find the index position of each maximum value in the image (return the two-dimensional coordinates).

[0081] Step 4: Perform average filtering calculation on the 16 neighborhoods of the image.

[0082] Use generic_filter to perform neighborhood average filtering on the image, size=8: specifies the neighborhood size as 8x8.

[0083] mode='nearest': The boundary pixels are processed by nearest neighbor interpolation. All positions of the first 8 maximum values ​​are traversed to extract the corresponding 16 neighborhood averages. Among the extracted 16 neighborhood averages, find the minimum and maximum values, replace min_mean with the parameter r1 of the piecewise function, and replace max_mean with r2. Define the output range of the piecewise function: s1=0 corresponds to r1, s2=250 corresponds to r2, and calculate the slopes of the three-segment function: k1: the slope of the first segment ([0, r1]), k2: the slope of the second segment ([r1, r2]), k3: the slope of the third segment ([r2, 256]). Use np.zeros to create a zero value matrix with the same size as the original image and data type np.uint8 (0-255) to save the processed pixel values. Traverse each pixel (i, j) of the image, select the corresponding piecewise function formula for linear transformation according to the range of pixel values: first segment: [0, r1], second segment: [r1, r2], third segment: [r2, 256], and store the transformation result in photoshape_img.

[0084] In summary, the method for contour enhancement of fluorescence in situ hybridization images based on illumination fusion and improved local binary patterns described in this application removes high-frequency noise (such as random noise) in the image through neighborhood mean filtering, retains the main structure and contour features of the cell, makes the image smoother and more natural, and provides a cleaner basis for subsequent analysis. The most prominent areas in the image (such as cell boundaries or high-contrast areas) are extracted, and the local characteristics of its neighborhood statistics (such as minimum and maximum values) are calculated to achieve the ability to dynamically adjust processing parameters. Finally, the overall contrast and local details of the image are enhanced by combining piecewise linear transformation, and the cell contours and structures are further highlighted by adapting to cell images with different brightness and contrast characteristics, providing high-quality visual input for analysis tasks such as tissue cell image segmentation, target detection and feature extraction.

[0085] Those skilled in the art will appreciate that the above are only preferred embodiments of the present invention, and the various embodiments of the present disclosure and / or the features described in the claims may be combined or combined in various ways, even if such combinations or combinations are not explicitly described in the present disclosure. It is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art may still modify the technical solutions described in the aforementioned embodiments, or perform equivalent substitutions on some of the technical features therein. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention shall be included in the protection scope of the present invention.

[0086] Although preferred embodiments of the present invention have been described, additional changes and modifications may be made to these embodiments by those skilled in the art once the basic inventive concepts are known. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention. Obviously, those skilled in the art may make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A method for enhancing the contour of fluorescence in situ hybridization images by illumination fusion and local binary, characterized in that: The method comprises the following steps: Step 1: Use the median filtering method to remove salt and pepper noise in the image, calculate the texture weight of the image through a linear equation, use the texture weight and smoothing function to analyze and process the illumination and texture features in the image to obtain a brightness image and a smoothed image, and perform grayscale processing on the brightness image and the smoothed image to obtain grayscale distribution information; Step 2: Based on the grayscale distribution information of the image, a local binary algorithm is used to select coefficients for extracting the contour feature image, and the image is enhanced to obtain a light-enhanced image; Step 3: Perform weighted addition of the contour feature image and the illumination enhancement image, and use a thermal map to obtain an image with a high energy gradient; Step 4, analyzing the eight neighborhoods around the eight maximum pixel values, setting breakpoints for the linear function according to the analysis results, determining the optimization range of the linear function for the image, identifying the area with high energy gradient through piecewise linear function and thermal map, and its high-frequency signal area is the contour area of ​​the fluorescence in situ hybridization tissue cell image; The method of calculating the texture weight of the image by the linear equation in step 1 is: obtaining by solving the constructed sparse matrix; The texture weight of the image in step 1 also includes the steps of calculating the texture weight of the image in the vertical direction difference and the horizontal direction difference; Calculate the vertical difference of the image, that is, the difference between each pixel and the next pixel, extend the vertical difference to the image boundary to form a periodic difference, calculate the horizontal difference of the image, process the boundary pixels, and apply one-dimensional smoothing filters to the horizontal and vertical differences respectively; The method for forming the periodic difference is: extending the difference in the vertical direction to the image boundary and connecting them in sequence from beginning to end to form a periodic difference.

2. The method for enhancing the edge of fluorescence in situ hybridization images by illumination fusion and local binary value according to claim 1, characterized in that: The coefficients in step 2 are achieved by choosing the 25% and 75% quantiles.

3. The method for enhancing the edge of fluorescence in situ hybridization images by illumination fusion and local binary value according to claim 1, characterized in that: The image enhancement processing in step 3 includes the steps of illumination smoothing, texture enhancement and brightness adjustment.

4. The method for enhancing the edge of fluorescence in situ hybridization images by illumination fusion and local binary value according to claim 1, characterized in that: Step 4 also includes the step of performing average filtering calculation on 16 neighborhoods of the image.

5. A computer device comprising a memory and a processor, characterized in that: A computer program is stored in the memory. When the processor runs the computer program stored in the memory, the processor executes the method according to any one of claims 1 to 4.

6. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 4 are implemented.

Images