Research On Overlapping Cell Segmentation Based On Sparse Edge Detection

Cervical Cancer is one of the most difficult problems in medical treatment.Screening for cervical cancer is of great significance in reducing morbidity and mortality.Early screening can effectively protect women's health.At present,cervical cell picture technology is a widely used detection and identification technology,which has the advantages of fast film production speed and high smear quality.In the automatic detection system of thin-layer cells,the automatic segmentation of cell images is the key and foundation of the whole system.Therefore,effectively segmenting overlapping cell images is the key to image classification and recognition,which is very important for improving the segmentation accuracy of the whole system.Therefore,this paper studies the overlapping image of cervical cells based on thin layer cell detection technology.Firstly,the image of the cervical cells is pre-processed,and the foreground object is modeled by using the Gaussian model,and then the ideal threshold is selected according to the function monotonicity to extract the foreground image of the cell.Image enhancement and denoising are performed on the cell image using morphological operations according to cell characteristics,eliminating impurities and complex backgrounds in the staining process.Secondly,an improved method based on the combination of block threshold and adaptive nucleus segmentation is proposed by analyzing the morphological characteristics of the nucleus.The method divides the nucleus into two parts: initial segmentation and re-segmentation.After the end of the segmentation,the cells are screened and classified,and then into the re-segmented part for secondary segmentation.Each part is improved by the characteristics of the nucleus,and classified and segmented according to the shape of the nucleus to achieve accurate segmentation,accuracy and segmentation effect of the nucleus.There has been a big improvement.Thirdly,a cell image segmentation method based on sparse edge detection model is proposed.The method divides the process of finding the cytoplasmic edge into two steps according to the characteristics of the cell: the strong edge and the weak edge,and the edge of the cell is represented by a series of sparse edge points.The model first locates the strong edge between the cell and the background image as the base contour,uses the edge-based dynamic search strategy to find the weak edge,and uses the obtained edge point as the base contour.Using the Snake active contour model to solve the function minimum value,the cell realism is approximated.The contours eventually reach the goal of dividing the cytoplasm.The designed method is simulated and tested to verify its accuracy and effectiveness.The experimental results show that the proposed method has better segmentation effect than the external force field GVF Snake algorithm and watershed algorithm,and has a great improvement in segmentation accuracy.