Polyp detection in wireless capsule endoscopy images using novel color texture features

Wireless capsule endoscopy (WCE) has been widely used nowadays to view the entire intestine. However, there still exists a challenging problem of it that over 55,000 images are produced during one examination for one patient. Analyzing such large number of images is very time-consuming and tedious for physicians, so we propose a computer aided diagnosis (CAD) strategy to automatically detect the polyps. Hopefully it will reduce this heavy burden for doctors. In this paper, we propose a novel opponent color feature integrating with LLE-based LBP texture feature. These two types of features are used to distinguish the polyp images from the normal ones using support vector machine (SVM). Experiments are performed on our current data set to verify the proposed scheme. The classification accuracy reaches 97% which shows that the proposed method is very efficient and promising.