Cell Nucleus Segmentation Based on Fuzzy Growing Snake

In this paper an automated cell nucleus segmentation method is proposed based on a novel fuzzy growing snake model for color cell images from esophageal smears. After estimating the ellipse parameters for each cell nucleus, the image point is mapped into several fuzzy domains which measure tinctorial uniformity and positional proximity of the pixel to the nucleus. Based on these fuzzy measures, a growing active contour model is developed to track the nucleus boundary. To help the model overcome local minimum, a self-adaptive growing energy is added to inflate the contour until it expands out of the nucleus. The utilization of ellipse estimation gives the model the strong edge reconstruction ability for the overlapped nucleus. Experiments show that the method has excellent segmentation results and stable performance.