Research On White Blood Cell Image Segmentation Based On Threshold And Watershed Algorithm

In this paper,algorithms such as color space conversion,threshold segmentation,maximum inter-class variance method,morphological reconstruction,and Watershed were used to study the segmentation of cytoplasm and nucleus in the two cases of non-adherent and adherent Wleucocytes.In this paper,two improved algorithms are proposed to segment the cytoplasm and nucleus of leukocytes in the pathological cell images,respectively,the threshold and watershed-based leukocyte segmentation algorithm and the improved watershed segmentation algorithm.The segmentation algorithm based on threshold and watershed is mainly used to segment the nucleus and cytoplasm of adherent and non-adherent leukocytes in the microscopic images mixed with background,erythrocyte,leukocyte plasm,and leukocyte nucleus.For the segmentation of the nucleus,this method firstly separates the general contour of the nucleus by threshold segmentation,then removes the isolated points by open reconstruction,and finally separates the nucleus by expansion operation to remove the irrelevant regions.The primary localization of leukocytes was realized by the separation of the nucleus.For the segmentation of cytoplasm,there are two cases.For non-adherent white blood cells,the threshold segmentation algorithm is directly used to segment the cytoplasm.Then,the gradient image was obtained by using Sobel,and the watershed transformation was performed according to the obtained gradient amplitude.Finally,the gray image was extracted by an open-close operation based on reconstruction to segment the cytoplasm.Using reconstruction-based open-close operation to segment cytoplasm can better remove image fragments without affecting the target image,and achieve a better segmentation effect.The improved watershed white blood cell segmentation algorithm is mainly used to segment the pathological white blood cells with adhesion phenomenon in the microscopic images mixed with background,red blood cells,leukocyte plasm,and leukocyte nucleus.For the segmentation of the nucleus,this method firstly converts the RGB color space of the cell image to the HSI color space and then extracts the H,S,and I components of the leukocyte image.Finally,the maximal inter-class variance method was used to extract the nuclei of the attached leukocytes from the S-component images.For the division of the cytoplasm,the method of S component image first,calculating the gradient image,based on Sobel operator with the refactoring operations to eliminate open operation details,then,the application of internal and external marking,make the minimum only appear in the internal and external markers,the watershed algorithm is utilized to extract adhesion and the white blood cells in the cytoplasm of white.The experimental verification shows that the white blood cell segmentation algorithm based on threshold and watershed can accurately segment the cytoplasm and nucleus from the lesion cell image with high efficiency.Compared with the traditional segmentation algorithm,the segmentation result of the improved watershed white blood cell segmentation algorithm is more accurate,not easily affected by lighting and dyeing environment,strong stability,short segmentation time,faster segmentation speed,and meet the real-time requirements of clinical diagnosis.