An Image Segmentation Method for Quasi-circular Immune Cells

Abstract

Aiming at the characteristic of actual quasi-circular immune cell images, this paper presents the method of quasicircular immune cell images segmentation based on Otsu threshold and thinning algorithm. The image is first converted color space form RGB to YIQ. Then the image is segmented by Otsu threshold algorithm. And then the erosion and dilation of morphological filter are used to process the image. Finally, the Zhang-Suen thinning algorithm is employed to extract the cell's skeleton, which is the center of the quasi-circular immune cell. According to the thinning times, we can obtain the radius value of the quasi-circular immune cell, and the overlapping quasi-circular immune cells are separated. Experimental results show this method works successfully in the segmentation of quasi-circular immune cell images.