An improved diffusion driven watershed algorithm for image segmentation of cells

The image segmentation is one of the most crucial steps in automated analysis of medical and biological images. The segmentation process allows for a detection of object contours. Due to specificity of imaging technique, a correct detection of cell contours is problematic because of the fuzzy and broken edges. Moreover, the cells are very often connected. The modified watershed algorithm based on the diffusion model presented in this paper has been successfully applied to segmentation of cells where the mentioned difficulties appear. The method was tested in contact endoscopy, a novel technique in the diagnosis of the larynx.