Segmenting multiple overlapping Nuclei in H&E stained breast cancer histopathology images based on an improved watershed

In histopathology images, there often exists several Nuclei overlapped with each other which causes difficulty to automatic nuclei segmentation. As we all know, watershed algorithm has been widely employed in image segmentation. But the limitation of watershed segmentation is sensitive to noise and can lead to serious over-segmentation. In this paper, we present an improved watershed transformation that incorporates opening-closing reconstruction and the distance transform with chamfer algorithm after color deconvolution, and H-minima . Unlike the classical watershed segmentation algorithm our improved method is able to resolve oversegmentation. The experiment results demonstrate our method successfully segment out each nuclei on breast cancer histology images, effectively address over-segmentation existed in traditional watershed segmentation , and preserve the original edges of each nuclei in the image completely.