Segmentation of tumors from endoscopic images using topological derivatives based on discrete approach

The growth and development of interdisciplinary research areas in the recent years has motivated the researchers in the area of image processing to apply the techniques from the Physics and Mathematics extensively in developing efficient image processing algorithms. One such application is use of topological derivatives for image segmentation. In this paper the cancer tumors from endoscopic images are segmented using the topological derivatives based on discrete approach. The topological derivatives quantify the sensitivity of a problem when the image domain is perturbed by the introduction of heterogeneity such as hole, inclusion, source term, etc. The topological derivatives are used here as a descent direction to minimize the associated cost function. The segmentation results based on the proposed methodology are very encouraging..