Using granulometry and watershed for breast tumor cells segmentation

In this paper, we present an efficient method based on mathematical morphology for segmenting microscopic breast cells images. The proposed approach consists on estimating the size cells distribution using granulometry. Drawing the spectrum will give us important information about different size distribution corresponding to the different cells present in the image. This information will be used in order to eliminate non tumor cells distinguished by their small size and irregular shape compared to tumor cells. The markers are then extracted using the maximum on the distance function of the resulted image. Finally, we apply the watershed algorithm. The proposed method is simple and efficient.