New features for wireless capsule endoscopy polyp detection

In this paper we present a new feature descriptor for automatic detection of frames with polyp in Wireless Capsule Endoscopy (WCE) images. The new approach is based on the fact that the polyp disease exhibits discriminating features when the WCE images are decomposed into different resolution levels. Hence we have made use of wavelet and emphasis feature extraction approaches. The 2-D discrete wavelet transform, dual tree complex wavelet transform, gabor wavelet transform and curvelet transform have been exploited to find out which one of them combined with probability distribution is suitable for polyp detection. Experiments were done on an augmented dataset and the results are satisfactory achieving 96% in term of performance.