Cervical cell nucleus multi-scale accurate segmentation method based on HSV channel difference features

Abstract

The application provides a cervical cell nucleus multi-scale accurate segmentation method based on an HSV channel difference feature, the method first converts an image to an HSV space and extracts a saturation S and a brightness V channel; then, adaptive enhanced difference feature channels are constructed by fusing a weighted difference value and a ratio relationship, the discrimination of the cell nucleus and the background is effectively improved, and the method has strong robustness to staining and illumination changes; subsequently, cell density sensing adaptive CLAHE and gradient guided adaptive median filtering are used for multi-scale enhancement and denoising, interference is suppressed while details are retained; in the segmentation stage, a hierarchical strategy of a global threshold value, a local adaptive threshold value and a morphological gradient enhancement is adopted, and adaptive morphological post-processing and screening are combined, so that accurate and complete segmentation of the cell nucleus in a complex scene is finally realized; the whole process of the application is based on clear image processing principles, does not require large-scale training data, has high calculation efficiency and is easy to integrate into an existing pathological information system.

Description

Technical Field

[0001] This invention relates to the field of image segmentation technology, and in particular to a method for precise multi-scale segmentation of cervical cell nuclei based on HSV channel difference features. Background Technology

[0002] Cervical cancer is the fourth most common cancer among women worldwide, and early screening is crucial for reducing mortality. Liquid-based thin-layer cytology (TCT) combined with Papanicolaou staining is currently the standard screening method, in which the morphological characteristics of the cell nucleus (such as size, shape, nucleocytoplasmic ratio, and staining intensity) are key evidence for diagnosing atypical cells and precancerous lesions.

[0003] With the widespread adoption of whole-slice digital pathology scanning technology, achieving precise segmentation of cell nuclei from massive amounts of high-resolution images has become a core bottleneck in improving the automation and accuracy of pathological diagnosis. After Papanicolaou staining, cell nuclei are stained deep blue-purple by hematoxylin (high saturation, low brightness), while cytoplasm is stained pink or orange by eosin (medium-low saturation, medium-high brightness). However, in practical applications, achieving precise segmentation faces multiple technical challenges:

[0004] Color interference and batch variation: Traditional RGB space segmentation methods are extremely sensitive to changes in lighting and staining batches, and the colors of the cell nucleus and cytoplasm overlap in the RGB channels, resulting in poor robustness of segmentation methods based on fixed thresholds.

[0005] Single feature utilization: Existing technologies mostly use a single channel in the HSV space in isolation (such as using only saturation S or brightness V), failing to systematically integrate the essential optical and physical characteristics of the cell nucleus of "high S and low V", resulting in insufficient feature differentiation.

[0006] Scale and morphological diversity: From normal small round nuclei to abnormal large nuclei, polymorphic nuclei, and lobed nuclei, their size and shape vary greatly, making it difficult for algorithms with fixed parameters to take into account all of them at the same time.

[0007] Blurred boundaries and severe adhesion: Dense cell areas often exhibit low boundary contrast, mutual adhesion, or even overlap, making edge- or region-based methods prone to undersegmentation or oversegmentation.

[0008] Noise and artifact interference: Optical noise and electronic noise during image acquisition, as well as impurities and bubbles generated during the staining process, can interfere with the extraction of the real target.

[0009] Existing solutions, such as those based on edge detection, region growing, or single-channel thresholding, have limited segmentation accuracy in complex scenarios. Deep learning methods, which rely on large-scale labeled data, suffer from poor interpretability, high computational resource requirements, and difficulty adapting to different staining protocols. Therefore, there is an urgent need for a precise cell nucleus segmentation technology that does not rely on massive amounts of labeled data, offers strong interpretability, and can adaptively address the aforementioned challenges. Summary of the Invention

[0010] This invention proposes a multi-scale precise segmentation method for cervical cell nuclei based on HSV channel differential features, which solves the problems of poor segmentation accuracy, reliance on massive labeled data, and poor interpretability of traditional cell nucleus segmentation methods.

[0011] The technical solution of this invention is implemented as follows:

[0012] This invention provides a method for precise multi-scale segmentation of cervical cell nuclei based on HSV channel differential characteristics, comprising the following steps:

[0013] S1, Image Acquisition and Multi-channel Feature Extraction: Acquire digital pathological images of cervical cells and convert them from RGB color space to HSV color space, and extract image data of saturation (S) and brightness (V) channels respectively.

[0014] S2, Adaptive Enhancement of Differential Feature Channel Construction: Based on the saturation S and brightness V channels, an enhanced differential feature channel is constructed by fusing weighted differences and ratios.

[0015] S3, Multi-scale adaptive contrast enhancement: The enhanced differential feature channels are subjected to cell density-aware adaptive CLAHE enhancement processing, and the enhanced image is filtered using an adaptive median filtering algorithm.

[0016] S4, Hierarchical adaptive threshold segmentation: The image after enhancement filtering is processed sequentially using global threshold segmentation and local adaptive threshold segmentation to generate a binary segmentation mask;

[0017] S5, Morphological post-processing: Adaptive morphological operations are performed on the binary segmentation mask to remove noise and fill holes, and connected region filtering based on morphological features is performed to obtain the final cell nucleus segmentation mask.

[0018] Specifically, in step S2, the enhanced differential feature channel Constructed using the following formula:

[0019] ;

[0020] in, The normalization function maps the input channel values ​​to the [0,1] interval; S and V are the image data for the saturation and brightness channels, respectively. To prevent small constants from being divided by zero, , , These are the weighting coefficients.

[0021] Furthermore, the weighting coefficients are calculated using the following formulas:

[0022] ;

[0023] ;

[0024] ;

[0025] ;

[0026] in, , These are the mean and standard deviation of the saturation channel S, respectively; , These are the maximum and minimum pixel values ​​for the luminance channel V, respectively; It is a constant; , These are the mean and standard deviation of the brightness channel V, respectively; The information entropy of the brightness channel V. This represents the average gradient value of the image pixels.

[0027] Specifically, in step S3, the cell density-sensing-based adaptive CLAHE enhancement processing method is as follows:

[0028] Estimating local cell density of an image The calculation formula is as follows:

[0029] ;

[0030] Where M×N is the size of the neighborhood window, Indicates the image at position Pixel value at that location, R is the density threshold; R is the neighborhood radius, defined as... The size of the square neighborhood centered on it; This is an indicator function; the function value is 1 when the condition within the parentheses is true, and 0 otherwise.

[0031] The CLAHE mesh size is dynamically adjusted based on local cell density, using the following formula:

[0032] ;

[0033] in, Indicates position The CLAHE grid size at that location.

[0034] Specifically, in step S3, the method for filtering the enhanced image using an adaptive median filtering algorithm includes:

[0035] Calculate the position of each pixel in the image Local gradient magnitude :

[0036] ;

[0037] in, , These represent the gradients of the image in the x and y directions, respectively.

[0038] Based on local gradient magnitude Dynamically select the window size for median filtering :

[0039] ;

[0040] , ;

[0041] in, , These represent the mean and standard deviation of the local gradient magnitudes of the image, respectively.

[0042] Specifically, in step S4, the global threshold segmentation method is as follows:

[0043] Global thresholding is performed based on the Otsu algorithm to obtain preliminary candidate regions for cell nuclei; the Otsu algorithm finds the optimal threshold T by maximizing the inter-class variance.

[0044] ;

[0045] in, This represents the inter-class variance when the threshold is T. , These represent the probabilities of foreground and background pixels after segmentation, respectively. , These represent the average pixel grayscale values ​​of the foreground and background after segmentation, respectively.

[0046] Furthermore, in step S4, the local adaptive threshold segmentation method is as follows:

[0047] Within the candidate cell nucleus region, each pixel is calculated based on local neighborhood statistical features. Local adaptive threshold Perform fine segmentation:

[0048] ;

[0049] in, Represented in pixels A local neighborhood window centered on the center, For neighborhood windows The total number of pixels within; For pixels The grayscale value; C is the adjustment parameter; pixels within the neighborhood window The grayscale value.

[0050] Further, in step S4, after the local adaptive threshold segmentation process, the boundary enhancement of the local threshold segmentation result is performed by combining the morphological gradient to generate a binary segmentation mask; the morphological gradient is defined as the difference between the result of the morphological dilation operation and the result of the morphological erosion operation on the image.

[0051] Specifically, step S5 includes the following steps:

[0052] The size of the structuring element used for opening operations is dynamically determined based on the average area of ​​cell nuclei in the current image, using the following formula:

[0053] ;

[0054] in, The size of the structuring element used for opening operations; The average area of ​​cell nuclei in the current image is given by k, where k is the scaling factor. This is the floor function;

[0055] Based on the shape characteristics of the cell nucleus, the corresponding structural element type is selected from elliptical, rectangular or cross-shaped structural elements, and the binary segmentation mask is opened to remove noise; then, a structural element suitable for the opening operation is selected to close the binary segmentation mask after the opening operation to fill the small holes inside the cell nucleus.

[0056] Identify all connected regions in the binary mask after morphological operations and filter them according to preset morphological criteria. The filtering criteria include: the area of ​​the connected region is within a preset range, the circularity is greater than a set threshold, and the convexity is greater than a set threshold. Regions that do not meet the filtering criteria are removed to obtain the final cell nucleus segmentation mask.

[0057] Preferably, the segmentation method further includes:

[0058] S6, Result Visualization: Extract the cell nucleus contour coordinates from the final cell nucleus segmentation mask; draw the contour coordinates with a preset high-contrast color and overlay them onto the original cervical cell color image to generate and output the cell nucleus contour marker image.

[0059] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0060] (1) By converting to HSV space and constructing enhanced features that fuse saturation and brightness, the present invention fundamentally separates the optical features of the cell nucleus and cytoplasm, reducing the sensitivity to staining and light changes; by fusing the weighted difference and ratio of S and V channels, a composite feature with significant distinguishability is constructed, overcoming the defect of insufficient information caused by relying on a single channel in traditional methods.

[0061] (2) By introducing weight coefficients that are dynamically calculated based on the global and local statistical characteristics of the image, the present invention enables the constructed enhanced difference features to adapt to the actual scenarios of different staining quality, lighting conditions and image content, so that the core features have excellent robustness to batch differences and complex background interference, and significantly improves the generalization ability and stability of the method on pathological images from different sources.

[0062] (3) By combining cell density-aware CLAHE with local gradient-guided adaptive median filtering, this invention achieves a dynamic balance between enhancement and smoothing. It can preserve fine structure and boundaries in dense cell areas, effectively enhance contrast in sparse areas, protect details in strong edge areas, and suppress noise in flat areas, providing high-quality intermediate images with prominent features and good noise suppression for subsequent segmentation.

[0063] (4) The Otsu global threshold → local statistical threshold → morphological gradient enhancement three-layer segmentation strategy adopted in this invention is closely coordinated with the subsequent adaptive opening and closing operation → multi-dimensional morphological screening post-processing process, which ensures that the segmentation process has both a global view and can make local fine adjustments, and can intelligently repair and purify the preliminary results. Finally, even in complex scenarios, it can still output cell nucleus segmentation results with accurate boundaries, complete morphology and biological characteristics. Attached Figure Description

[0064] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0065] Figure 1This is a flowchart illustrating a method for precise multi-scale segmentation of cervical cell nuclei based on HSV channel differential characteristics, as described in this invention.

[0066] Figure 2 This is a color image of the original cervical cells obtained in an embodiment of the present invention.

[0067] Figure 3 This is the saturation-brightness difference feature image obtained after step S2 in this embodiment of the invention.

[0068] Figure 4 This is the enhanced image obtained after step S3 in this embodiment of the invention.

[0069] Figure 5 This is a cell nucleus outline marker image obtained through step S6 in an embodiment of the present invention. Detailed Implementation

[0070] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0071] Reference Figure 1 This invention provides a method for precise multi-scale segmentation of cervical cell nuclei based on HSV channel differential characteristics, comprising the following steps:

[0072] Step S1, Image Acquisition and Multi-channel Feature Extraction:

[0073] Obtain digital pathological images of cervical cells stained with Papanicolaou (e.g.) Figure 2 As shown in the figure, the image data of the two independent channels of saturation (S) and lightness (V) are extracted from the RGB color space and converted from the HSV color space.

[0074] Step S2, Adaptive Enhancement of Differential Feature Channel Construction:

[0075] Based on the saturation (S) and brightness (V) channels, an enhanced difference feature channel is constructed by fusing weighted differences and ratios. The enhanced saturation-brightness difference feature image is shown below. Figure 3 As shown;

[0076] The enhanced differential feature channel Constructed using the following formula:

[0077] ;

[0078] in, The normalization function maps the input channel values ​​to the [0,1] interval; S and V are the image data for the saturation and brightness channels, respectively. To prevent division by zero, a small constant (usually set to 1) is used. , , These are weighting coefficients dynamically calculated based on the statistical characteristics of the image.

[0079] The weighting coefficient , , The results are obtained using the following formulas:

[0080] ;

[0081] ;

[0082] ;

[0083] ;

[0084] in, , These are the pixel mean and standard deviation of the saturation channel S, respectively; , These are the maximum and minimum pixel values ​​for the luminance channel V, respectively; It is a constant; , These are the pixel mean and standard deviation of the luminance channel V, respectively; The information entropy of the brightness channel V. This represents the average gradient value of the image pixels.

[0085] Step S3, Multi-scale adaptive contrast enhancement:

[0086] For the enhanced differential feature channel The image undergoes cell density-aware adaptive CLAHE (histogram equalization) enhancement processing, followed by adaptive median filtering. The enhanced image after enhancement filtering is shown below. Figure 4 As shown;

[0087] The cell density-sensing adaptive CLAHE enhancement method is as follows:

[0088] 1) Estimate the local cell density of the image The calculation formula is as follows:

[0089] ;

[0090] Among them, local cell density M×N is the size of the neighborhood window (default is 15×15). Indicates the image at position Pixel value at that location, The density threshold is adaptively calculated based on image content; R is the neighborhood radius, defined by... The size of the square neighborhood centered on it; This is an indicator function; the function value is 1 when the condition within the parentheses is true, and 0 otherwise.

[0091] 2) The CLAHE mesh size is dynamically adjusted based on the local cell density, using the following formula:

[0092] ;

[0093] in, Indicates position The CLAHE grid size is used at each location; for densely populated areas, a smaller grid size (e.g., 4×4) is used; for sparsely populated areas, a larger grid size (e.g., 8×8) is used.

[0094] Methods for filtering enhanced images using adaptive median filtering algorithms include:

[0095] 1) Calculate the position of each pixel in the image Local gradient magnitude :

[0096] ;

[0097] in, , These represent the gradients of the image in the x and y directions, respectively.

[0098] 2) Based on local gradient magnitude Dynamically select the window size for median filtering :

[0099] ;

[0100] , ;

[0101] in, , These represent the mean and standard deviation of the local gradient magnitudes of the image, respectively.

[0102] Step S4, hierarchical adaptive threshold segmentation:

[0103] The image after enhancement filtering is processed sequentially using global thresholding and local adaptive thresholding, and the local thresholding results are enhanced by combining morphological gradients to generate a binary segmentation mask. The morphological gradient is defined as the difference between the result of morphological dilation and morphological erosion, which can effectively highlight the boundaries of objects.

[0104] The global threshold segmentation method is as follows:

[0105] Global thresholding is performed based on the Otsu algorithm to obtain preliminary candidate regions for cell nuclei; the Otsu algorithm finds the optimal threshold T by maximizing the inter-class variance.

[0106] ;

[0107] in, This represents the inter-class variance when the threshold is T. , These represent the probabilities of foreground and background pixels after segmentation, respectively. , These represent the average pixel grayscale values ​​of the foreground and background after segmentation, respectively.

[0108] The local adaptive threshold segmentation method is as follows:

[0109] Within the candidate cell nucleus region, each pixel is calculated based on local neighborhood statistical features. Local adaptive threshold Perform fine segmentation:

[0110] ;

[0111] in, Represented in pixels A local neighborhood window centered on the center, For neighborhood windows The total number of pixels within; For pixels The grayscale value; C is an adjustment parameter (the empirical value is -1 to 2, and it is 1 in this embodiment); pixels within the neighborhood window The grayscale value.

[0112] Step S5, Morphological Post-processing: Adaptive morphological operations are performed on the binary segmentation mask to remove noise and fill holes (first, opening operations are performed with small structuring elements to remove pepper noise, and then closing operations are performed with appropriate structuring elements to fill small holes in the nucleus), and connected region filtering based on morphological features is performed to obtain the final cell nucleus segmentation mask, specifically including the following steps:

[0113] 1) The size of the structuring element used for opening operations is dynamically determined based on the average area of ​​the cell nuclei in the current image. The calculation formula is as follows:

[0114] ;

[0115] in, The size of the structuring element used for opening operations; The average area (in pixels) of cell nuclei in the current image is given by k, which is a scaling factor (empirical value is 0.15). This is the floor function;

[0116] 2) Based on the shape characteristics of the cell nucleus, select the corresponding structural element type from elliptical, rectangular, or cross-shaped structural elements, and perform an opening operation on the binary segmentation mask to remove noise; then select a structural element suitable for the opening operation to perform a closing operation on the binary segmentation mask after the opening operation to fill the small holes inside the cell nucleus; the correspondence between the selected structural element type and the shape characteristics of the cell nucleus is shown in Table 1 below:

[0117] Table 1. Correspondence between selected structural element types and cell nucleus shape characteristics.

[0118] Structural Element Type Applicable Scenarios Shape characteristics Ellipse Most round / oval cell nuclei Isotropic smoothness Rectangle (RECT) A well-defined, regularly shaped cell nucleus Maintain right angle characteristics CROSS Separating adherent cell nuclei Maintain a slender structure

[0119] Based on the shape characteristics of the cell nucleus, refer to Table 1 above to select appropriate structural elements for opening or closing operations.

[0120] 3) Identify all connected regions in the binary mask after morphological operations, and filter them according to preset morphological criteria. The filtering criteria include: the area of ​​the connected region is within a preset range (e.g., 20-500 pixels), the circularity is greater than a set threshold (e.g., 0.3), and the convexity is greater than a set threshold (e.g., 0.7). Regions that do not meet the filtering criteria are removed to obtain the final cell nucleus segmentation mask.

[0121] Step S6, result visualization:

[0122] Extract cell nucleus contour coordinates from the final cell nucleus segmentation mask; draw the contour coordinates with a preset high-contrast color (such as green) and overlay them onto the original cervical cell color image to generate and output an intuitive cell nucleus contour marker image, such as... Figure 5 As shown.

[0123] The segmentation method of the present invention has the following beneficial effects:

[0124] Strong robustness: The D_enhanced channel, constructed by fusing weighted differences and ratios, fundamentally amplifies the inherent color characteristics of the cell nucleus and exhibits excellent robustness to staining differences and light changes.

[0125] High segmentation accuracy: Multi-scale adaptive preprocessing and hierarchical thresholding strategy effectively address cell size variations and local contrast changes, improving the accuracy of boundary segmentation.

[0126] Reliable positional accuracy: The cell nucleus contours on the original image are marked according to the contour coordinates, which ensures the accurate correspondence between biological features (contours) and image space, solves the problem of result offset in traditional methods, and lays a reliable foundation for subsequent quantitative analysis.

[0127] It is practical and interpretable: the entire process is based on clear image processing principles, requires no large-scale training data, has high computational efficiency, and is easy to integrate into existing pathology information systems (LIS / PACS) or computer-aided diagnostic (CAD) platforms to assist pathologists in conducting rapid and objective analysis.

[0128] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for precise multi-scale segmentation of cervical cell nuclei based on HSV channel differential characteristics, characterized in that, Includes the following steps: S1, Image Acquisition and Multi-channel Feature Extraction: Acquire digital pathological images of cervical cells and convert them from RGB color space to HSV color space, and extract image data of saturation (S) and brightness (V) channels respectively. S2, Adaptive Enhancement of Differential Feature Channel Construction: Based on the saturation S and brightness V channels, an enhanced differential feature channel is constructed by fusing weighted differences and ratios. S3, Multi-scale adaptive contrast enhancement: The enhanced differential feature channels are subjected to cell density-aware adaptive CLAHE enhancement processing, and the enhanced image is filtered using an adaptive median filtering algorithm. S4, Hierarchical adaptive threshold segmentation: The image after enhancement filtering is processed sequentially using global threshold segmentation and local adaptive threshold segmentation to generate a binary segmentation mask; S5, Morphological post-processing: Adaptive morphological operations are performed on the binary segmentation mask to remove noise and fill holes, and connected region filtering based on morphological features is performed to obtain the final cell nucleus segmentation mask.

2. The method for precise multi-scale segmentation of cervical cell nuclei based on HSV channel differential characteristics as described in claim 1, characterized in that, In step S2, the enhanced differential feature channel Constructed using the following formula: ； in, The normalization function maps the input channel values ​​to the [0,1] interval; S and V are the image data for the saturation and brightness channels, respectively. To prevent small constants from being divided by zero, , , These are the weighting coefficients.

3. The method for precise multi-scale segmentation of cervical cell nuclei based on HSV channel differential characteristics as described in claim 2, characterized in that, The weighting coefficients are calculated using the following formulas: ； ； ； ； in, , These are the mean and standard deviation of the saturation channel S, respectively; , These are the maximum and minimum pixel values ​​for the luminance channel V, respectively; It is a constant; , These are the mean and standard deviation of the brightness channel V, respectively; The information entropy of the brightness channel V. This represents the average gradient value of the image pixels.

4. The method for precise multi-scale segmentation of cervical cell nuclei based on HSV channel differential characteristics as described in claim 1, characterized in that, In step S3, the cell density-sensing adaptive CLAHE enhancement processing method is as follows: Estimating local cell density of an image The calculation formula is as follows: ； Where M×N is the size of the neighborhood window, Indicates the image at position Pixel value at that location, R is the density threshold; R is the neighborhood radius, defined as... The size of the square neighborhood centered on it; This is an indicator function; the function value is 1 when the condition within the parentheses is true, and 0 otherwise. The CLAHE mesh size is dynamically adjusted based on local cell density, using the following formula: ； in, Indicates position The CLAHE grid size at that location.

5. The method for precise multi-scale segmentation of cervical cell nuclei based on HSV channel differential characteristics as described in claim 1, characterized in that, In step S3, the method for filtering the enhanced image using an adaptive median filtering algorithm includes: Calculate the position of each pixel in the image Local gradient magnitude : ； in, , These represent the gradients of the image in the x and y directions, respectively. Based on local gradient magnitude Dynamically select the window size for median filtering : ； ， ； in, , These represent the mean and standard deviation of the local gradient magnitudes of the image, respectively.

6. The method for precise multi-scale segmentation of cervical cell nuclei based on HSV channel differential characteristics as described in claim 1, characterized in that, In step S4, the global threshold segmentation method is as follows: Global thresholding is performed based on the Otsu algorithm to obtain preliminary candidate regions for cell nuclei; the Otsu algorithm finds the optimal threshold T by maximizing the inter-class variance. ； in, This represents the inter-class variance when the threshold is T. , These represent the probabilities of foreground and background pixels after segmentation, respectively. , These represent the average pixel grayscale values ​​of the foreground and background after segmentation, respectively.

7. The method for precise multi-scale segmentation of cervical cell nuclei based on HSV channel differential characteristics as described in claim 6, characterized in that, In step S4, the local adaptive threshold segmentation method is as follows: Within the candidate cell nucleus region, each pixel is calculated based on local neighborhood statistical features. Local adaptive threshold Perform fine segmentation: ； in, Represented in pixels A local neighborhood window centered on the center, For neighborhood windows The total number of pixels within; For pixels The grayscale value; C is the adjustment parameter; pixels within the neighborhood window The grayscale value.

8. The method for precise multi-scale segmentation of cervical cell nuclei based on HSV channel differential characteristics as described in claim 1, characterized in that, In step S4, after local adaptive threshold segmentation, the boundary of the local threshold segmentation result is enhanced by combining morphological gradient to generate a binary segmentation mask; the morphological gradient is defined as the difference between the result of the morphological dilation operation and the result of the morphological erosion operation on the image.

9. The method for precise multi-scale segmentation of cervical cell nuclei based on HSV channel differential characteristics as described in claim 1, characterized in that, Step S5 specifically includes the following steps: The size of the structuring element used for opening operations is dynamically determined based on the average area of ​​cell nuclei in the current image, using the following formula: ； in, The size of the structuring element used for opening operations; The average area of ​​cell nuclei in the current image is given by k, where k is the scaling factor. This is the floor function; Based on the shape characteristics of the cell nucleus, the corresponding structural element type is selected from elliptical, rectangular or cross-shaped structural elements, and the binary segmentation mask is opened to remove noise; then, a structural element suitable for the opening operation is selected to close the binary segmentation mask after the opening operation to fill the small holes inside the cell nucleus. Identify all connected regions in the binary mask after morphological operations and filter them according to preset morphological criteria. The filtering criteria include: the area of ​​the connected region is within a preset range, the circularity is greater than a set threshold, and the convexity is greater than a set threshold. Regions that do not meet the filtering criteria are removed to obtain the final cell nucleus segmentation mask.

10. The method for precise multi-scale segmentation of cervical cell nuclei based on HSV channel differential characteristics as described in claim 1, characterized in that, Also includes: S6, Result Visualization: Extract the cell nucleus contour coordinates from the final cell nucleus segmentation mask; draw the contour coordinates with a preset high-contrast color and overlay them onto the original cervical cell color image to generate and output the cell nucleus contour marker image.

Images