Sparsity analysis of endoscopy images

Sparsity analysis of images is important to understand the image characteristic and its possible potential in applications. In this paper, normal images and endoscopy image are investigated for their sparsity using K-SVD algorithm that finds a dictionary basis with minimal number of non-zero coefficients in the transformed domain to have minimal prediction error. The results show that the endoscopy image has lower prediction error and lower number of non-zero coefficients in the transformed domain. This indicates the fact that one can develop a better endoscopy image encoder with better prediction mechanism, and a better endoscopy image decoder with better error concealment method to recover data contaminated by the noise, both based on the idea of sparsity.