Segmentation of Papanicolaou smear images

A segmentation method for images from Papanicolaou smears is proposed that will locate the cellular boundary based on the stability of the perimeter of the cell and will locate the nucleus based on the thresholding of a smoothed histogram. It emphasizes the selection of the thresholds for nuclear and cytoplasmic boundary tracking. The problem of touching cells was handled by using texture information in the touching area as well as shape information from the cellular boundary. The proposed method, as applied to 11 scenes containing 19 cells of different classes, worked satisfactorily without any human interactions.