Peripheral blood red blood cell image segmentation method based on adaptive and edge enhancement

Abstract

This invention relates to the field of in vitro diagnostics and discloses a method for segmenting peripheral blood red blood cell images based on adaptive and edge enhancement. The method includes: acquiring an image of red blood cells to be segmented; performing contour enhancement on the image of red blood cells to be segmented to obtain a target red blood cell image; calculating the gray-level histogram of the target red blood cell image and extracting the envelope line from the gray-level histogram; determining two main peaks and the valley between the two main peaks on the envelope line using a maximum value filtering method; using the pixel value corresponding to the valley as a threshold to perform threshold segmentation on the target red blood cell image to obtain a binary image; performing segmentation point detection and matching on the binary image, connecting the matched segmentation points to segment individual red blood cells; and performing ellipse fitting on the individual red blood cells to optimize the red blood cell contour. The method disclosed in this application solves the problem of effectively segmenting red blood cells, effectively separating adhered red blood cells and hypochromic red blood cells.

Description

Technical Field

[0001] This invention relates to the field of in vitro diagnostics, and in particular to a method for segmenting peripheral blood erythrocyte images based on adaptive and edge enhancement. Background Technology

[0002] Peripheral blood cell morphology analysis in clinical practice mainly examines the morphology of red blood cells, white blood cells, and platelets for changes to determine the presence of hematological disorders. If the red blood cells appear large, it may indicate myelodysplastic syndrome, hemolytic anemia, or megaloblastic anemia. If the red blood cells are small, it may indicate thalassemia or iron deficiency anemia. Before identifying red blood cell morphology, the peripheral blood cell morphology analyzer needs to segment the red blood cells in the image.

[0003] In related technologies, methods for segmenting red blood cells in red blood cell images include thresholding, watershed segmentation, edge segmentation, and deep learning image segmentation. Thresholding segmentation is further divided into global thresholding, Otsu thresholding, iterative thresholding, and bimodal thresholding. Edge segmentation methods include Sobel, Canny, Laplace, and unsharp mask (USM) for edge extraction. These methods have advantages in different scenarios. For example, thresholding works well when there is a large and stable difference between foreground and background colors, but performs poorly in scenarios with uneven lighting and complex backgrounds. Edge segmentation works well for images with clear outlines, but edges are often discontinuous, requiring complex post-processing and being sensitive to noise. Watershed segmentation also segments based on edge information and can produce continuous boundaries, but it is prone to oversegmentation and is also sensitive to noise. Deep learning image segmentation achieves good segmentation results, but requires a large amount of labeled data and is dependent on hardware such as graphics cards, otherwise computation is time-consuming.

[0004] In a fully automated blood cell morphology analyzer, it is necessary to extract features such as size, color, shape, and contents of peripheral blood red blood cells. These features (especially cell size) usually require first dividing individual red blood cells to obtain a red blood cell mask, and then extracting the corresponding cell features.

[0005] Currently, there is an urgent need for an optimized red blood cell segmentation method for fully automated blood cell morphology analyzers. Summary of the Invention

[0006] The purpose of this invention is to provide at least one method for segmenting peripheral blood red blood cells based on adaptive and edge enhancement, which can at least solve the problem of effectively segmenting red blood cells and at least achieve the effect of effectively segmenting clump-together red blood cells and hypochromic red blood cells.

[0007] To address the aforementioned technical problems, at least one embodiment of this application provides a peripheral blood red blood cell image segmentation method based on adaptive and edge enhancement, comprising: acquiring a red blood cell image to be segmented; performing contour enhancement on the red blood cell image to be segmented to obtain a target red blood cell image; calculating a gray-level histogram of the target red blood cell image and extracting the envelope line in the gray-level histogram; determining two main peaks and the valley between the two main peaks on the envelope line using a maximum value filtering method; using the pixel value corresponding to the valley as a threshold to perform threshold segmentation on the target red blood cell image to obtain a binary image; performing segmentation point detection and matching on the binary image, connecting the matched segmentation points to segment independent red blood cells; and performing ellipse fitting on the independent red blood cells to optimize the red blood cell contour.

[0008] At least one embodiment of this application also provides a peripheral blood red blood cell image segmentation device based on adaptive and edge enhancement, comprising: an image acquisition module for acquiring an image of red blood cells to be segmented; a contour enhancement module for enhancing the contour of the image of red blood cells to be segmented to obtain a target red blood cell image; a histogram calculation module for calculating a gray-level histogram of the target red blood cell image and extracting the envelope line in the gray-level histogram; a valley determination module for determining two main peaks on the envelope line and the valley between the two main peaks using a maximum value filtering method; a threshold segmentation module for using the pixel value corresponding to the valley as a threshold to perform threshold segmentation on the target red blood cell image to obtain a binary image; a segmentation point detection and matching module for detecting and matching segmentation points in the binary image and connecting the matched segmentation points to segment independent red blood cells; and an ellipse fitting module for performing ellipse fitting on the independent red blood cells to optimize the red blood cell contour.

[0009] At least one embodiment of this application also provides an electronic device, including: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the above-described adaptive and edge-enhanced peripheral blood red blood cell image segmentation method.

[0010] At least one embodiment of this application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method for peripheral blood red blood cell image segmentation based on adaptive and edge enhancement.

[0011] At least one embodiment of this application also provides a blood cell morphology analyzer, comprising: a smear support module for supporting a peripheral blood smear; an imaging module for capturing images of the peripheral blood smear; a movement module for driving the peripheral blood smear to move horizontally or vertically relative to the imaging module, so that a region of interest in the peripheral blood smear enters the field of view of the imaging module; an image processing module communicatively connected to the imaging module, for at least acquiring independent red blood cell images; and an image analysis module for identifying red blood cell types based on the independent red blood cell images; wherein the image processing module is configured to: acquire a first image captured by the imaging module through a low-power objective lens, and determine the red blood cell imaging region of the peripheral blood smear based on the first image; acquire the... The imaging module captures a second image of red blood cells in the red blood cell imaging area using a high-magnification objective lens; acquires an image of red blood cells to be segmented based on the second image; performs contour enhancement on the image of red blood cells to be segmented to obtain a target red blood cell image; calculates the gray-level histogram of the target red blood cell image and extracts the envelope line from the gray-level histogram; determines the two main peaks on the envelope line and the troughs between the two main peaks using the maximum value filtering method; uses the pixel value corresponding to the trough as a threshold to perform threshold segmentation on the target red blood cell image to obtain a binary image; performs segmentation point detection and matching on the binary image, connects the matched segmentation points to segment independent red blood cells; and performs ellipse fitting on the independent red blood cells to optimize the red blood cell contour and obtain an independent red blood cell image.

[0012] The peripheral blood erythrocyte image segmentation method based on adaptive edge enhancement provided in this application enhances the contour of the erythrocyte image to be segmented to obtain the target erythrocyte image. It strengthens the bimodal histogram feature to effectively compensate for the strong dependence of the bimodal histogram method on the bimodal curve, improving the accuracy of subsequent threshold segmentation and segmentation point detection and matching. The gray-level histogram of the target erythrocyte image is calculated, and the envelope in the gray-level histogram is extracted. The two main peaks and the troughs between the two main peaks on the envelope are determined by the maximum value filtering method. The pixel values ​​corresponding to the troughs are used as thresholds to perform threshold segmentation on the target erythrocyte image to obtain a binary image. This effectively resists noise and multi-peak interference, thus adapting more flexibly to peripheral blood erythrocyte images. Furthermore, segmentation point detection and matching are performed on the binary image. Matched segmentation points are connected to segment individual erythrocytes, effectively separating adhered erythrocytes into individual erythrocytes, thereby improving the accuracy of erythrocyte morphology analysis. Furthermore, ellipse fitting is performed on individual red blood cells to optimize their outlines. This can effectively solve the problem of unclear or broken edges caused by the light-stained areas of low-pigment red blood cells, making it impossible to extract complete red blood cells. By optimizing the edges of red blood cells with abnormal edges in the binary image through ellipse fitting, it is possible to avoid the problem of unclear or broken red blood cell edges caused by the expansion of the light-stained area to the edge, which would lead to inaccurate calculation of features such as cell area, cell diameter, cell shape, and diameter of the light-stained area.

[0013] In some optional embodiments, obtaining the red blood cell image to be segmented includes: obtaining an RGB image of the red blood cells; converting the RGB image of the red blood cells into a grayscale image; and denoising the grayscale image to obtain the red blood cell image to be segmented. By denoising the grayscale image, the interference of noise on contour enhancement and grayscale histogram calculation can be reduced, thereby obtaining a more continuous contour and a more accurate threshold.

[0014] In some optional embodiments, the contour enhancement of the red blood cell image to be segmented is performed using an enhanced convolution kernel filtering formula to obtain the target red blood cell image; the enhanced convolution kernel filtering formula is as follows:

[0015]

[0016] In the formula, dst represents the output target red blood cell image, src represents the pixel matrix, GaussianBlur() represents the Gaussian blur function, dilate() represents the dilation function, and erode() represents the erosion function.

[0017] By using the above-mentioned enhanced convolution kernel filtering formula, the contour enhancement of the red blood cell image to be segmented is performed. It focuses more on strong contour information and is not sensitive to noise, which greatly improves the accuracy of subsequent threshold segmentation and segmentation point detection and matching.

[0018] In some optional embodiments, ellipse fitting is performed on the individual red blood cells to optimize their contours. This includes: performing ellipse fitting on the individual red blood cells; if the ratio of the area of ​​the fitted red blood cell to that before fitting exceeds a set threshold, the edge of the red blood cell is determined to be broken; for the red blood cells with broken edges, a rotational scan is performed based on the center of the ellipse to determine two tangent points at the broken edge of the red blood cell; using these two tangent points and the center of the ellipse, ellipse fitting is performed again to connect the broken edge in an arc shape, forming a complete red blood cell contour. By using the above ellipse fitting method to optimize and complete the contours of low-chromic red blood cells with broken edges, low-chromic red blood cells can be accurately and effectively extracted, which is beneficial to improving the accuracy of subsequent red blood cell morphology recognition.

[0019] In some optional embodiments, after fitting the individual red blood cells to an ellipse to optimize the red blood cell contour, the method further includes: for red blood cells with broken edges, increasing the distance from the moving point to the two foci while keeping the foci and center of the ellipse unchanged, and performing a rotational scan again based on the center of the ellipse; if a region satisfying a preset condition is detected between two tangent points, adjusting the arc connecting the broken edges to expand the area of ​​the red blood cell, the preset condition including an area exceeding a preset area and a pixel value exceeding a preset pixel threshold. Based on the distance from the moving point to the two foci of the ellipse, the ellipse coverage is expanded by increasing the preset distance, and a rotational scan is performed again based on the center of the ellipse. If a region exceeding a preset area and a preset pixel threshold is detected between the two tangent points of the missing edge, the missing edge is expanded and connected to verify and correct the missing edge, ensuring more accurate segmented cell boundaries.

[0020] In some optional embodiments, segmentation point detection and matching are performed on the binary image, and the matched segmentation points are connected to segment individual red blood cells. This includes: extracting the outer contour of red blood cells from the binary image; detecting concave points on the outer contour of the red blood cells; performing non-maximum suppression based on all detected concave points to retain valid concave points; calculating the direction vector between the current concave point and its adjacent concave points; if the direction of the direction vector is consistent with the direction of the current concave point, then the current concave point and its adjacent concave points are matched; determining the matched concave points as segmentation points, and connecting the segmentation points to segment individual red blood cells. Combining contour enhancement adaptive threshold segmentation and segmentation point detection matching can effectively segment tightly adhered red blood cells, thereby improving the accuracy of red blood cell morphology analysis. Attached Figure Description

[0021] One or more embodiments are illustrated by way of example with reference to the accompanying drawings, and these illustrative descriptions do not constitute a limitation on the embodiments.

[0022] Figure 1 This is a flowchart of the peripheral blood erythrocyte image segmentation method based on adaptive and edge enhancement provided in the embodiments of this application;

[0023] Figure 2 This is a schematic diagram showing the relative positional relationships between different regions in a peripheral blood smear provided in the embodiments of this application;

[0024] Figure 3 This is a schematic diagram of the movement trajectory of red blood cells captured by a low-power objective lens according to an embodiment of this application;

[0025] Figure 4 This is a schematic diagram of an RGB image of red blood cells provided in an embodiment of this application;

[0026] Figure 5 This is a schematic diagram of central red blood cell segmentation using the Otsu method provided in an embodiment of this application;

[0027] Figure 6 This is a schematic diagram of the red blood cell segmentation results provided in an embodiment of this application;

[0028] Figure 7 This is a schematic diagram of red blood cells with broken edges due to the expansion of the lightly stained area provided in the embodiments of this application;

[0029] Figure 8 This is a schematic diagram of hypochromic erythrocytes provided in an embodiment of this application;

[0030] Figure 9 This is a schematic diagram of the result after two ellipse fittings provided in the embodiments of this application;

[0031] Figure 10 This is a schematic diagram illustrating the results of each step provided in the embodiments of this application;

[0032] Figure 11 This is a schematic diagram of the ellipse fitting process results provided in the embodiments of this application;

[0033] Figure 12 This is a schematic diagram of a peripheral blood erythrocyte image segmentation device based on adaptive and edge enhancement provided in an embodiment of this application. Detailed Implementation

[0034] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the various embodiments of this application will be described in detail below with reference to the accompanying drawings. However, those skilled in the art will understand that many technical details have been provided in the various embodiments of this application to help readers better understand this application. However, the technical solutions claimed in this application can be implemented even without these technical details and various changes and modifications based on the following embodiments. The division of the various embodiments below is for the convenience of description and should not constitute any limitation on the specific implementation of this application. The various embodiments can be combined with and referenced by each other without contradiction.

[0035] To facilitate understanding of the embodiments of this application, the relevant content on peripheral blood cell morphology analysis will be introduced first.

[0036] Peripheral blood cell morphology analysis in clinical practice mainly examines the morphology of red blood cells, white blood cells, and platelets for changes to determine the presence of hematological disorders. If the red blood cells appear large, it may indicate myelodysplastic syndrome, hemolytic anemia, or megaloblastic anemia. If the red blood cells are small, it may indicate thalassemia or iron deficiency anemia. Before identifying red blood cell morphology, the peripheral blood cell morphology analyzer needs to segment the red blood cells in the image.

[0037] In related technologies, methods for segmenting red blood cells in red blood cell images include thresholding, watershed segmentation, edge segmentation, and deep learning image segmentation. Thresholding, in particular, is further divided into global thresholding, Otsu thresholding, iterative thresholding, and bimodal thresholding. Edge segmentation methods include Sobel operators, Kenny operators, Laplacian operators, and unsharpened masks for edge extraction.

[0038] These methods each have their advantages in different scenarios. For example, thresholding works well when the difference between foreground and background colors is large and stable, but performs poorly in scenarios with uneven lighting and complex backgrounds. Edge segmentation works well for images with clear outlines, but the edges are often discontinuous, requiring more complex post-processing to complete the segmentation, and it is also sensitive to noise. Watershed segmentation also segments based on edge information and can produce continuous boundaries, but it is prone to oversegmentation and is also sensitive to noise. Deep learning image segmentation methods have good segmentation results, but require a large amount of labeled data and depend on hardware such as graphics cards, otherwise the computation time is too long.

[0039] In fully automated hematology analyzers, it is necessary to extract features such as size, color, shape, and contents from peripheral blood erythrocytes. These features (especially cell size) usually require first segmenting individual erythrocytes to obtain a erythrocyte mask before extracting the corresponding cell features. During erythrocyte segmentation, there may be cases where erythrocytes adhere to each other, or even where the edges of pale-stained areas are unclear or broken.

[0040] To address the technical challenges of effective red blood cell segmentation in cases of adhesion between red blood cells and edge defects in the pale-stained areas of hypopigmented red blood cells, this invention proposes a peripheral blood red blood cell image segmentation method based on adaptive and edge enhancement. The implementation details of this embodiment of the peripheral blood red blood cell image segmentation method based on adaptive and edge enhancement are described below. The following content is provided for ease of understanding and is not essential for implementing this solution.

[0041] Example 1:

[0042] The peripheral blood erythrocyte image segmentation method based on adaptive and edge enhancement in this embodiment can be applied to electronic devices with communication, computing, and data storage capabilities. Its specific process can be as follows: Figure 1 As shown, it includes:

[0043] Step 101: Obtain the image of the red blood cells to be segmented.

[0044] In a specific implementation, obtaining the image of the red blood cells to be segmented can include:

[0045] Step 101a: Obtain the RGB image of red blood cells.

[0046] In a specific implementation, the smear-carrying module carries the peripheral blood smear, and the moving module drives the peripheral blood smear to move horizontally or vertically relative to the imaging module so that the region of interest in the peripheral blood smear enters the field of view of the imaging module. The imaging module takes a first image through a low-power objective lens (e.g., a 10x objective lens) to search and determine the red blood cell imaging region of the peripheral blood smear based on the first image. The imaging module takes a second image of the red blood cells in the red blood cell imaging region through a high-power objective lens (e.g., a 50x objective lens, a 100x objective lens, etc.). There may be multiple second images taken under the high-power objective lens (e.g., a 50x or 100x objective lens in conjunction with a 0.5x conversion lens), and there may be overlap between adjacent second images. Therefore, all the second images taken under the high-power objective lens are stitched together to obtain the red blood cell RGB image obtained in step 101a.

[0047] In some cases, the red blood cell imaging area includes a monolayer of cells in the body tail region of a peripheral blood smear. The relative positions of the various regions in the peripheral blood smear are as follows: Figure 2 As shown, the movement trajectory of red blood cells is captured by a low-power objective lens. Figure 3 As shown, the red blood cell imaging area was determined by searching under a 10x objective lens. Multiple second images were captured under a 100x objective lens, with overlap between adjacent images. All the second images from the 100x objective lens were stitched together to obtain the RGB image of the red blood cells in the imaging area, as shown below. Figure 4 As shown.

[0048] Step 101b: Convert the RGB image of red blood cells to a grayscale image.

[0049] Step 101c: Denoise the grayscale image to obtain the red blood cell image to be segmented.

[0050] In some cases, Gaussian filtering, a linear smoothing filter, can be used to denoise grayscale images. Using Gaussian filtering to denoise grayscale images can reduce the interference of noise on contour enhancement and grayscale histogram calculation, thereby obtaining more continuous contours and more accurate thresholds.

[0051] In other examples, grayscale images can be denoised using one of the following methods: mean filtering, median filtering, bilateral filtering, guided filtering, or other denoising techniques.

[0052] Step 102: Perform contour enhancement on the red blood cell image to be segmented to obtain the target red blood cell image.

[0053] In some examples, the enhanced convolution kernel filtering formula is used to enhance the contours of the red blood cell image to be segmented, thereby obtaining the target red blood cell image; the enhanced convolution kernel filtering formula is as follows:

[0054]

[0055] In the formula, dst represents the output target red blood cell image, src represents the pixel matrix, GaussianBlur() represents the Gaussian blur function, dilate() represents the dilation function, and erode() represents the erosion function.

[0056] By using the above-mentioned enhanced convolution kernel filtering formula, the contour enhancement of the red blood cell image to be segmented is performed. It focuses more on strong contour information and is not sensitive to noise, which greatly improves the accuracy of subsequent threshold segmentation and segmentation point detection and matching.

[0057] In other examples, contour enhancement can employ any of the following: morphological gradient, unsharpened mask, Laplacian, LOG (Laplace of Gaussian), or DOG (Difference of Gaussians).

[0058] Step 103: Calculate the grayscale histogram of the target red blood cell image and extract the envelope from the grayscale histogram.

[0059] In the actual implementation, after extracting the envelope from the grayscale histogram, a smoothing process is also performed to remove fluctuations and noise.

[0060] Step 104: Determine the two main peaks on the envelope and the trough between the two main peaks using the maximum value filtering method.

[0061] In practical implementation, the two main peaks refer to the peak corresponding to the foreground and the peak corresponding to the background.

[0062] Step 105: Use the pixel value corresponding to the valley as the threshold to perform threshold segmentation on the target red blood cell image to obtain a binary image.

[0063] Because red blood cells are characterized by relatively uniform overall color, thresholding segmentation usually yields good results. However, peripheral blood red blood cells often exhibit issues such as uneven staining (too dark or too light), blurred image scanning, and cell adhesion. Therefore, thresholding methods like global thresholding and Otsu thresholding often fail to achieve satisfactory segmentation results. Figure 5 As shown, although the Otsu threshold method can roughly segment red blood cells, it is easily affected by noise and cannot handle cell adhesion problems.

[0064] The bimodal histogram method is an adaptive thresholding segmentation method that automatically finds the threshold based on the grayscale histogram of each image. Compared to the Otsu thresholding method, it is more dependent on the bimodal histogram feature of the image, but it is easier to extend and has better flexibility. This embodiment improves upon the bimodal histogram method by using maximum value filtering to find two main peaks, thereby effectively resisting noise and interference from multiple peaks (small peaks), thus more flexibly adapting to the segmentation needs of peripheral blood erythrocyte images. Specifically, firstly, edge enhancement strengthens the bimodal histogram feature, effectively compensating for the strong dependence of the bimodal histogram method on the bimodal curve. Secondly, maximum value filtering is used to find the peak value of the histogram curve. By adjusting the "coverage range" of the peaks, the influence of multiple peaks and noise can be effectively avoided. The combination of edge enhancement and maximum value filtering can effectively determine the segmentation threshold and obtain better segmentation results.

[0065] Step 106: Perform segmentation point detection and matching on the binary image, and connect the matched segmentation points to separate individual red blood cells.

[0066] Since threshold segmentation only considers the grayscale information of the image and ignores geometric and spatial information, this embodiment performs segmentation point detection and matching on the segmented binary mask in order to independently segment tightly adhered red blood cells. Specifically: the outer contour of the red blood cells is extracted from the binary (mask) image, and concave points are detected on the outer contour of the red blood cells; after detecting all concave points, non-maximum suppression is performed based on all detected concave points to retain valid concave points; the direction vector between the current concave point and its adjacent concave points is calculated; if the direction vector is consistent with the direction of the current concave point, the current concave point and its adjacent concave points are matched; the matched concave points are determined as segmentation points, and the segmentation points are connected to segment independent red blood cells. Figure 6 As shown, combining contour enhancement adaptive threshold segmentation and segmentation point detection matching can effectively separate central red blood cells, thereby improving the accuracy of red blood cell morphology analysis.

[0067] Step 107: Perform ellipse fitting on individual red blood cells to optimize the red blood cell profile.

[0068] like Figure 6 and Figure 7 As shown, in reality, hypochromic erythrocytes often exist. The pale areas of hypochromic erythrocytes are very close to the background color of the image. By calculating the area ratio of the pale areas to the erythrocytes, it is possible to identify hypochromic erythrocytes. Figure 8 The low-chromic red blood cells within the red box often have enlarged pale-stained areas, sometimes even extending to the edges, resulting in unclear or broken edges. In such cases, the segmentation algorithm cannot extract the entire red blood cell, leading to inaccurate calculations of cell area, cell diameter, cell shape, and pale-stained area diameter. This embodiment uses the individual red blood cells obtained in steps 101-106 for ellipse fitting to identify red blood cells with enlarged pale-stained areas that result in broken edges (such as...). Figure 7 Contour completion (as shown) can effectively and accurately segment individual red blood cells, thereby improving the accuracy of subsequent red blood cell morphology recognition.

[0069] In the specific implementation, individual red blood cells undergo ellipse fitting to optimize their contours, including:

[0070] Step 107a: Perform ellipse fitting on individual red blood cells;

[0071] Step 107b: If the ratio of the area of ​​the red blood cells after fitting to that before fitting exceeds a set threshold, the edge of the red blood cell is determined to be broken.

[0072] Step 107c: For red blood cells with broken edges, perform a rotational scan based on the center of the ellipse to determine two tangent points at the broken edge of the red blood cell. Use these two tangent points and the center of the ellipse to perform ellipse fitting again, and connect the broken edge in an arc shape to form a complete red blood cell outline.

[0073] By using the ellipse fitting method described above, the outline of hypochromic erythrocytes with broken edges can be optimized and completed, enabling the accurate and effective extraction of hypochromic erythrocytes, which is beneficial to improving the accuracy of subsequent erythrocyte morphology recognition.

[0074] In one example, the results of steps 101 to 107 of the method in this embodiment are as follows: Figure 10 As shown. The process of segmenting red blood cells with broken edges and then fitting an ellipse is as follows: Figure 11 As shown.

[0075] This embodiment provides a peripheral blood erythrocyte image segmentation method based on adaptive edge enhancement. It enhances the contour of the erythrocyte image to be segmented to obtain the target erythrocyte image, strengthening the bimodal histogram feature to effectively compensate for the strong dependence of the bimodal histogram method on the bimodal curve, thus improving the accuracy of subsequent threshold segmentation and segmentation point detection and matching. The gray-level histogram of the target erythrocyte image is calculated, and the envelope in the gray-level histogram is extracted. The two main peaks and the troughs between the two main peaks are determined using the maximum value filtering method. The pixel values ​​corresponding to the troughs are used as thresholds to perform threshold segmentation on the target erythrocyte image, obtaining a binary image. This effectively resists noise and multi-peak interference, thus adapting more flexibly to peripheral blood erythrocyte images. Furthermore, segmentation point detection and matching are performed on the binary image. Connecting the matched segmentation points separates individual erythrocytes, effectively separating adhered erythrocytes into individual erythrocytes, thereby improving the accuracy of erythrocyte morphology analysis. Furthermore, ellipse fitting is performed on individual red blood cells to optimize their outlines. This can effectively solve the problem of unclear or broken edges caused by the light-stained areas of low-pigment red blood cells, making it impossible to extract complete red blood cells. By optimizing the edges of red blood cells with abnormal edges in the binary image through ellipse fitting, it is possible to avoid the problem of unclear or broken red blood cell edges caused by the expansion of the light-stained area to the edge, which would lead to inaccurate calculation of features such as cell area, cell diameter, cell shape, and diameter of the light-stained area.

[0076] In some embodiments, after ellipse fitting of individual red blood cells to optimize red blood cell contours, the method further includes:

[0077] Step 108: For red blood cells with broken edges, increase the distance from the moving point to the two foci while keeping the foci and center of the ellipse unchanged, and perform a rotation scan again based on the center of the ellipse; if a region that meets the preset conditions is detected between the two tangent points, adjust the arc connecting the broken edge to expand the area of ​​the red blood cell. The preset conditions include the area exceeding the preset area and the pixel value exceeding the preset pixel threshold.

[0078] In this embodiment, a preset distance is added to the distance from the moving point of the ellipse to the two foci to expand the coverage of the ellipse. The ellipse is rotated and scanned again based on the center of the ellipse. If a region with a preset pixel threshold exceeding the preset area is detected between the two tangent points of the missing edge, the missing edge is expanded and connected to verify and correct the missing edge, ensuring that the segmented cell boundaries are more accurate.

[0079] by Figure 7 Taking the red blood cells on the left as an example, after step 107, line 1 is fitted; after step 108, line 2 is fitted. The fitting results are as follows: Figure 9 As shown, by expanding the area of ​​the ellipse and fitting it again, the resulting red blood cell outline is more accurate and reliable.

[0080] The method in this embodiment has the following advantages:

[0081] First, the segmentation threshold can be adjusted flexibly and adaptively, thereby improving the robustness of red blood cell image segmentation. Compared with existing adaptive thresholding methods, such as the Otsu method, this method can adjust the segmentation threshold more flexibly according to the characteristics of red blood cell images, thus achieving stable and accurate segmentation results.

[0082] Second, by combining contour enhancement, threshold segmentation, and segmentation point detection algorithms, the segmentation ability of adhering red blood cells is improved, while continuous and stable segmentation contours are generated.

[0083] Third, by detecting the broken edge points through ellipse fitting, the complete edge of the broken hypochromic erythrocyte can be restored.

[0084] Example 2:

[0085] Another embodiment of this application relates to a peripheral blood erythrocyte image segmentation device based on adaptive and edge enhancement. The implementation details of the erythrocyte image segmentation device of this embodiment are described below. The following implementation details are provided for ease of understanding and are not essential for implementing this solution. A schematic diagram of the erythrocyte image segmentation device of this embodiment can be seen as follows: Figure 12 As shown, it includes an image acquisition module 201, a contour enhancement module 202, a histogram calculation module 203, a valley determination module 204, a threshold segmentation module 205, a segmentation point detection and matching module 206, and an ellipse fitting module 207.

[0086] Image acquisition module 201 is used to acquire images of red blood cells to be segmented;

[0087] The contour enhancement module 202 is used to enhance the contour of the red blood cell image to be segmented to obtain the target red blood cell image;

[0088] The histogram calculation module 203 is used to calculate the grayscale histogram of the target red blood cell image and extract the envelope line from the grayscale histogram. In the specific implementation, after extracting the envelope line from the grayscale histogram, smoothing processing is also performed to remove fluctuations and noise.

[0089] The trough determination module 204 is used to determine the two main peaks on the envelope and the trough between the two main peaks by using the maximum value filtering method; the two main peaks refer to the peak corresponding to the foreground and the peak corresponding to the background.

[0090] The threshold segmentation module 205 is used to perform threshold segmentation on the target red blood cell image by using the pixel value corresponding to the valley as the threshold to obtain a binary image.

[0091] The segmentation point detection and matching module 206 is used to detect and match segmentation points in a binary image, and connect the matched segmentation points to separate individual red blood cells.

[0092] Ellipse fitting module 207 is used to perform ellipse fitting on individual red blood cells to optimize the red blood cell contour.

[0093] In some embodiments, obtaining the red blood cell image to be segmented may include: obtaining a red blood cell RGB image; converting the red blood cell RGB image into a grayscale image; and denoising the grayscale image to obtain the red blood cell image to be segmented.

[0094] In the specific implementation, the smear carrying module carries the peripheral blood smear, and the moving module drives the peripheral blood smear to move horizontally or vertically relative to the imaging module so that the region of interest in the peripheral blood smear enters the field of view of the imaging module. The imaging module takes a first image through a low-power objective lens (e.g., a 10x objective lens) to search and determine the red blood cell imaging region of the peripheral blood smear based on the first image. The imaging module takes a second image of the red blood cells in the red blood cell imaging region through a high-power objective lens (e.g., a 50x or 100x objective lens in conjunction with a 0.5x conversion lens). There may be multiple second images taken under the high-power objective lens (e.g., a 100x objective lens), and there may be overlap between adjacent second images. Therefore, all the second images taken under the high-power objective lens are stitched together to obtain the red blood cell RGB image.

[0095] In some cases, the red blood cell imaging area includes a monolayer of cells in the body tail region of a peripheral blood smear. The relative positions of the various regions in the peripheral blood smear are as follows: Figure 2 As shown, the movement trajectory of red blood cells is captured by a low-power objective lens. Figure 3 As shown, the red blood cell imaging area was determined by searching under a 10x objective lens. Multiple second images were captured under a 100x objective lens, with overlap between adjacent images. All the second images from the 100x objective lens were stitched together to obtain the RGB image of the red blood cells in the imaging area, as shown below. Figure 4 As shown.

[0096] In some cases, Gaussian filtering, a linear smoothing filter, can be used to denoise grayscale images. Using Gaussian filtering reduces noise interference with contour enhancement and grayscale histogram calculation, resulting in more continuous contours and more accurate thresholds. In other cases, grayscale image denoising can employ one of the following methods: mean filtering, median filtering, bilateral filtering, guided filtering, or other denoising techniques.

[0097] In some examples, the enhanced convolution kernel filtering formula is used to enhance the contours of the red blood cell image to be segmented, thereby obtaining the target red blood cell image; the enhanced convolution kernel filtering formula is as follows:

[0098]

[0099] In the formula, dst represents the output target red blood cell image, src represents the pixel matrix, GaussianBlur() represents the Gaussian blur function, dilate() represents the dilation function, and erode() represents the erosion function.

[0100] By using the above-mentioned enhanced convolution kernel filtering formula, the contour enhancement of the red blood cell image to be segmented is performed. It focuses more on strong contour information and is not sensitive to noise, which greatly improves the accuracy of subsequent threshold segmentation and segmentation point detection and matching.

[0101] In other examples, contour enhancement can employ any of the following: morphological gradient, unsharpened mask, Laplacian, LOG (Laplace of Gaussian), or DOG (Difference of Gaussians).

[0102] Because red blood cells are characterized by relatively uniform overall color, thresholding segmentation usually yields good results. However, peripheral blood red blood cells often exhibit issues such as uneven staining (too dark or too light), blurred image scanning, and cell adhesion. Therefore, thresholding methods like global thresholding and Otsu thresholding often fail to achieve satisfactory segmentation results. Figure 5 As shown, although the Otsu threshold method can roughly segment red blood cells, it is easily affected by noise and cannot handle cell adhesion problems.

[0103] The bimodal histogram method is an adaptive thresholding segmentation method that automatically finds the threshold based on the grayscale histogram of each image. Compared to the Otsu thresholding method, it is more dependent on the bimodal histogram feature of the image, but it is easier to extend and has better flexibility. This embodiment improves upon the bimodal histogram method by using maximum value filtering to find two main peaks, thereby effectively resisting noise and interference from multiple peaks (small peaks), thus more flexibly adapting to the segmentation needs of peripheral blood erythrocyte images. Specifically, firstly, edge enhancement strengthens the bimodal histogram feature, effectively compensating for the strong dependence of the bimodal histogram method on the bimodal curve. Secondly, maximum value filtering is used to find the peak value of the histogram curve. By adjusting the "coverage range" of the peaks, the influence of multiple peaks and noise can be effectively avoided. The combination of edge enhancement and maximum value filtering can effectively determine the segmentation threshold and obtain better segmentation results.

[0104] Since threshold segmentation only considers the grayscale information of the image and ignores geometric and spatial information, this embodiment performs segmentation point detection and matching on the segmented binary mask in order to independently segment tightly adhered red blood cells. Specifically: the outer contour of the red blood cells is extracted from the binary (mask) image, and concave points are detected on the outer contour of the red blood cells; after detecting all concave points, non-maximum suppression is performed based on all detected concave points to retain valid concave points; the direction vector between the current concave point and its adjacent concave points is calculated; if the direction vector is consistent with the direction of the current concave point, the current concave point and its adjacent concave points are matched; the matched concave points are determined as segmentation points, and the segmentation points are connected to segment independent red blood cells. Figure 6 As shown, combining contour enhancement adaptive threshold segmentation and segmentation point detection matching can effectively separate central red blood cells, thereby improving the accuracy of red blood cell morphology analysis.

[0105] like Figure 6 and Figure 7 As shown, in reality, hypochromic erythrocytes often exist. The pale areas of hypochromic erythrocytes are very close to the background color of the image. By calculating the area ratio of the pale areas to the erythrocytes, it is possible to identify hypochromic erythrocytes. Figure 8 The red boxes within the hypochromic red blood cells highlight the issue of their pale stained areas. These pale stained areas often expand, sometimes even extending to the edges, resulting in unclear or broken edges. In such cases, the segmentation algorithm cannot extract the entire red blood cell, leading to inaccurate calculations of cell area, diameter, shape, and pale stained area diameter. This embodiment utilizes the segmented individual red blood cells for ellipse fitting to identify red blood cells with expanded pale stained areas that result in broken edges (such as...). Figure 7 Contour completion (as shown) can effectively and accurately segment individual red blood cells, thereby improving the accuracy of subsequent red blood cell morphology recognition.

[0106] In the specific implementation, ellipse fitting is performed on individual red blood cells to optimize the red blood cell contour. This includes: performing ellipse fitting on individual red blood cells; if the ratio of the red blood cell area after fitting to that before fitting exceeds a set threshold, the edge of the red blood cell is determined to be broken; for red blood cells with broken edges, a rotational scan is performed based on the center of the ellipse to determine two tangent points at the broken edge of the red blood cell, and ellipse fitting is performed again using these two tangent points and the center of the ellipse to connect the broken edge in an arc shape to form a complete red blood cell contour.

[0107] By using the ellipse fitting method described above, the outline of hypochromic erythrocytes with broken edges can be optimized and completed, enabling the accurate and effective extraction of hypochromic erythrocytes, which is beneficial to improving the accuracy of subsequent erythrocyte morphology recognition.

[0108] The peripheral blood erythrocyte image segmentation device based on adaptive and edge enhancement provided in this embodiment enhances the contour of the erythrocyte image to be segmented to obtain the target erythrocyte image. It strengthens the bimodal histogram feature to effectively compensate for the strong dependence of the bimodal histogram method on the bimodal curve, improving the accuracy of subsequent threshold segmentation and segmentation point detection and matching. The gray-level histogram of the target erythrocyte image is calculated, and the envelope in the gray-level histogram is extracted. The two main peaks and the troughs between the two main peaks are determined by the maximum value filtering method. The pixel values ​​corresponding to the troughs are used as thresholds to perform threshold segmentation on the target erythrocyte image to obtain a binary image. This effectively resists noise and multi-peak interference, thus adapting more flexibly to peripheral blood erythrocyte images. Furthermore, segmentation point detection and matching are performed on the binary image. Matched segmentation points are connected to segment individual erythrocytes, effectively separating adhered erythrocytes into individual erythrocytes, thereby improving the accuracy of erythrocyte morphology analysis. Furthermore, ellipse fitting is performed on red blood cells to optimize their outlines. This can effectively solve the problem of unclear or broken edges caused by the light-stained areas of low-pigment red blood cells, making it impossible to extract complete red blood cells. By using ellipse fitting to optimize the edges of red blood cells with abnormal edges in the binary image, it can avoid the problem of unclear or broken red blood cell edges caused by the expansion of the light-stained area to the edge, which would lead to inaccurate calculation of features such as cell area, cell diameter, cell shape, and diameter of the light-stained area.

[0109] In some embodiments, the ellipse fitting module 207 is further configured to: after performing ellipse fitting on individual red blood cells to optimize the red blood cell contour, for red blood cells with broken edges, increase the distance from the moving point to the two foci while keeping the foci and center of the ellipse unchanged, and perform a rotation scan again based on the center of the ellipse; if a region satisfying a preset condition is detected between two tangent points, adjust the arc connecting the broken edge to expand the area of ​​the red blood cell, the preset condition including the area exceeding a preset area and the pixel value exceeding a preset pixel threshold.

[0110] In this embodiment, a preset distance is added to the distance from the moving point of the ellipse to the two foci to expand the coverage of the ellipse. The ellipse is rotated and scanned again based on the center of the ellipse. If a region with a preset pixel threshold exceeding the preset area is detected between the two tangent points of the missing edge, the missing edge is expanded and connected to verify and correct the missing edge, ensuring that the segmented cell boundaries are more accurate.

[0111] by Figure 7 Taking the red blood cells on the left as an example, after step 107, line 1 is fitted; after step 108, line 2 is fitted. The fitting results are as follows: Figure 9 As shown, by expanding the area of ​​the ellipse and fitting it again, the resulting red blood cell outline is more accurate and reliable.

[0112] The device of this embodiment has the following advantages:

[0113] First, the segmentation threshold can be adjusted flexibly and adaptively, thereby improving the robustness of red blood cell image segmentation. Compared with existing adaptive thresholding methods, such as the Otsu method, this device can adjust the segmentation threshold more flexibly according to the characteristics of red blood cell images, thus achieving stable and accurate segmentation results.

[0114] Second, by combining contour enhancement, threshold segmentation, and segmentation point detection algorithms, the segmentation ability of adhering red blood cells is improved, while continuous and stable segmentation contours are generated.

[0115] Third, by detecting the broken edge points through ellipse fitting, the complete edge of the broken hypochromic erythrocyte can be restored.

[0116] It is worth mentioning that all modules involved in this embodiment are logical modules. In practical applications, a logical unit can be a physical unit, a part of a physical unit, or a combination of multiple physical units. Furthermore, to highlight the innovative aspects of this application, this embodiment does not introduce units that are not closely related to solving the technical problems proposed in this application; however, this does not mean that other units are absent in this embodiment.

[0117] Example 3:

[0118] Another embodiment of this application relates to an electronic device, including: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the peripheral blood red blood cell image segmentation method based on adaptive and edge enhancement in the above embodiments.

[0119] The memory and processor are connected via a bus, which can include any number of interconnecting buses and bridges, connecting various circuits of one or more processors and memories. The bus can also connect various other circuits, such as peripheral devices, voltage regulators, and power management circuits, which are well known in the art and will not be described further herein. The bus interface provides an interface between the bus and the transceiver. The transceiver can be a single element or multiple elements, such as multiple receivers and transmitters, providing a unit for communicating with various other devices over a transmission medium. Data processed by the processor is transmitted over the wireless medium via an antenna, which further receives data and transmits it to the processor.

[0120] The processor manages the bus and general processing, and also provides various functions, including timing, peripheral interfaces, voltage regulation, power management, and other control functions. Memory is used to store data used by the processor during operation.

[0121] Example 4:

[0122] Another embodiment of this application relates to a computer-readable storage medium storing a computer program. When executed by a processor, the computer program implements the above-described embodiment of the peripheral blood erythrocyte image segmentation method based on adaptive and edge enhancement.

[0123] That is, those skilled in the art will understand that all or part of the steps in the methods of the above embodiments can be implemented by a program instructing related hardware. This program is stored in a storage medium and includes several instructions to cause a device (which may be a microcontroller, chip, etc.) or processor to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0124] Example 5:

[0125] Another embodiment of this application relates to a blood cell morphology analyzer, comprising:

[0126] Smear carrier module, used to hold peripheral blood smears;

[0127] The imaging module is used to capture images of peripheral blood smears;

[0128] The moving module is used to drive the peripheral blood smear to move horizontally or vertically relative to the imaging module so that the region of interest in the peripheral blood smear enters the field of view of the imaging module.

[0129] An image processing module, which is communicatively connected to the imaging module, is used at least to acquire independent red blood cell images;

[0130] An image analysis module, based on independent red blood cell images, is used at least to identify red blood cell types;

[0131] The image processing module is configured to: acquire a first image captured by the imaging module using a low-power objective lens, and determine the red blood cell imaging region of the peripheral blood smear based on the first image; acquire a second image captured by the imaging module using a high-power objective lens targeting the red blood cells in the red blood cell imaging region; acquire the red blood cell image to be segmented based on the second image; perform contour enhancement on the red blood cell image to be segmented to obtain the target red blood cell image; calculate the gray-level histogram of the target red blood cell image and extract the envelope line from the gray-level histogram; determine the two main peaks and the valleys between the two main peaks on the envelope line using the maximum value filtering method; perform threshold segmentation on the target red blood cell image using the pixel values ​​corresponding to the valleys as thresholds to obtain a binary image; perform segmentation point detection and matching on the binary image, connect the matched segmentation points to segment independent red blood cells; and perform ellipse fitting on the independent red blood cells to optimize the red blood cell contours and obtain independent red blood cell images.

[0132] The specific implementation of the image processing module can be found in the aforementioned embodiments, and will not be repeated in this embodiment.

[0133] Those skilled in the art will understand that the above embodiments are specific embodiments for implementing this application, and in practical applications, various changes can be made to them in form and detail without departing from the spirit and scope of this application.

Claims

1. A peripheral blood erythrocyte image segmentation method based on adaptive and edge enhancement, characterized in that, include: Obtain the image of the red blood cells to be segmented; Contour enhancement is performed on the red blood cell image to be segmented to obtain the target red blood cell image; Calculate the grayscale histogram of the target red blood cell image and extract the envelope from the grayscale histogram; The two main peaks on the envelope and the trough between the two main peaks are determined by the maximum value filtering method. The pixel value corresponding to the valley is used as a threshold to perform threshold segmentation on the target red blood cell image to obtain a binary image. The binary image is segmented by detection and matching of segments, and the matched segments are connected to separate individual red blood cells. Ellipse fitting was performed on the individual red blood cells to optimize the red blood cell contour; Ellipse fitting is performed on the individual red blood cells to optimize their contours, including: Ellipse fitting was performed on the individual red blood cells; If the ratio of the area of ​​the red blood cells after fitting to that before fitting exceeds a set threshold, the edge of the red blood cell is determined to be broken. For red blood cells with broken edges, a rotational scan is performed based on the center of the ellipse to determine two tangent points at the broken edge of the red blood cell. Using these two tangent points and the center of the ellipse, an ellipse fitting is performed again to connect the broken edge in an arc shape, forming a complete red blood cell outline. After performing ellipse fitting on the individual red blood cells to optimize the red blood cell contours, the method further includes: For red blood cells with broken edges, while keeping the foci and center of the ellipse unchanged, the distance from the moving point to the two foci is increased, and a rotational scan is performed again based on the center of the ellipse; If a region satisfying a preset condition is detected between two tangent points, the arc connecting the edge break is adjusted to increase the area of ​​the red blood cell. The preset condition includes an area exceeding a preset area and a pixel value exceeding a preset pixel threshold.

2. The peripheral blood erythrocyte image segmentation method based on adaptive and edge enhancement according to claim 1, characterized in that, The process of acquiring the red blood cell image to be segmented includes: Obtain RGB images of red blood cells; Convert the red blood cell RGB image to a grayscale image; The grayscale image is denoised to obtain the red blood cell image to be segmented.

3. The peripheral blood erythrocyte image segmentation method based on adaptive and edge enhancement according to claim 1, characterized in that, The contour enhancement of the red blood cell image to be segmented is performed using the enhanced convolution kernel filtering formula to obtain the target red blood cell image. The formula for calculating the enhanced convolution kernel filter is as follows: In the formula, dst represents the output target red blood cell image, src represents the pixel matrix, GaussianBlur() represents the Gaussian blur function, dilate() represents the dilation function, and erode() represents the erosion function.

4. The peripheral blood erythrocyte image segmentation method based on adaptive and edge enhancement according to claim 1, characterized in that, The binary image is subjected to segmentation point detection and matching, and the matched segmentation points are connected to separate individual red blood cells, including: Extract the outer contour of red blood cells from the binary image, and perform concave point detection on the outer contour of red blood cells; Non-maximum suppression is performed based on all detected concave points to preserve valid concave points; Calculate the direction vector between the current concave point and its adjacent concave points; If the direction of the direction vector is consistent with the direction of the current concave point, then the current concave point and its adjacent concave points are matched. The matched concave points are identified as dividing points, and the dividing points are connected to separate individual red blood cells.

5. A peripheral blood erythrocyte image segmentation device based on adaptive and edge enhancement, wherein the method described in any one of claims 1-4 is characterized in that, include: The image acquisition module is used to acquire images of red blood cells to be segmented; A contour enhancement module is used to enhance the contour of the red blood cell image to be segmented to obtain a target red blood cell image; The histogram calculation module is used to calculate the gray-level histogram of the target red blood cell image and extract the envelope line from the gray-level histogram. The valley determination module is used to determine the two main peaks on the envelope and the valley between the two main peaks by using the maximum value filtering method. The threshold segmentation module is used to perform threshold segmentation on the target red blood cell image by using the pixel value corresponding to the valley as a threshold to obtain a binary image. The segmentation point detection and matching module is used to detect and match segmentation points in the binary image, and connect the matched segmentation points to separate independent red blood cells. An ellipse fitting module is used to perform ellipse fitting on the individual red blood cells to optimize the red blood cell contours.

6. An electronic device, characterized in that, include: At least one processor; as well as, A memory communicatively connected to the at least one processor; wherein, The memory stores instructions executable by the at least one processor, which, when executed by the at least one processor, enables the at least one processor to perform the peripheral blood red blood cell image segmentation method based on any one of claims 1 to 4.

7. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the peripheral blood red blood cell image segmentation method based on adaptive and edge enhancement as described in any one of claims 1 to 4.

8. A blood cell morphology analyzer, based on the method described in any one of claims 1-4, characterized in that, include: Smear carrier module, used to hold peripheral blood smears; An imaging module is used to capture images of the peripheral blood smear; A moving module is used to drive the peripheral blood smear to move horizontally or vertically relative to the imaging module, so that the region of interest in the peripheral blood smear enters the field of view of the imaging module; An image processing module, which is communicatively connected to the imaging module, is used at least to acquire independent red blood cell images; The image analysis module, based on the independent red blood cell images, is used at least to identify red blood cell types; The image processing module is configured as follows: The imaging module acquires a first image captured by a low-power objective lens, and the red blood cell imaging area of ​​the peripheral blood smear is determined based on the first image. The imaging module acquires a second image of the red blood cells in the red blood cell imaging area using a high-power objective lens; Obtain the red blood cell image to be segmented based on the second image; Contour enhancement is performed on the red blood cell image to be segmented to obtain the target red blood cell image; Calculate the grayscale histogram of the target red blood cell image and extract the envelope from the grayscale histogram; The two main peaks on the envelope and the trough between the two main peaks are determined by the maximum value filtering method. The pixel value corresponding to the valley is used as a threshold to perform threshold segmentation on the target red blood cell image to obtain a binary image. The binary image is segmented by detection and matching of segments, and the matched segments are connected to separate individual red blood cells. Ellipse fitting is performed on the individual red blood cells to optimize their contours and obtain individual red blood cell images.

Images