Abstract:Traditional clustering method had obvious errors in processing immunohistochemical digital image. In order to improve the accuracy of cells classification and counting

a new immunohistochemical digital image processing method was proposed in this paper

which normalized the image in LAB color space

and then convert it to HSV color space. This method uses the maximum between-cluster variation weighted clustering to classify different types of cells in the image

and uses morphological methods to segment and count the cells of different types respectively. Experiment demonstrates that the method can use the color information effectively

on the premise of maintaining the high noise tolerant capability

the method can use the color information effectively to realize the classification of cells in immunohistochemical digital image

and accomplish the segmentation and count accurately.