We propose an automatic and unsupervised method for the segmentation of colonic polyps for in vivo Narrow-Band-Imaging (NBI) data, during optical colonoscopy, aiming at the prevention of colon cancer. The method is based on the Chan & Vese segmentation model and involves the sum of different Wasserstein distances, relying on histograms of suitable image descriptors, such as the intensity, texture, scale and orientation.
This presentation is part of Contributed Presentation “CP2 - Contributed session 2”