Automatic Segmentation of Polyps in Endoscopic Image Using Level-Set Formulation

Abstract

Automatic segmentation of colorectal polyps from endoscopic images forms an interesting challenge in computer vision. The method proposed in this paper intends to segment colorectal polyp (abnormal) regions from normal regions from a given endoscopic image. Due to lack of any regular texture patterns in this kind of images and apparent visual similarity in background and foreground pixels, conventional texture feature extraction and classification methods do not always yield good results. Hence, active contour based method has been explored to automatically segment out probable abnormal region(s). Our aim is to automatically detect the probable polyp region(s) and then verify the results with respect to the ground truth. Due to lack of very definitive edge criteria along the boundaries of a polyp, we used “active contour without edges” instead of classical active contour.