Enhanced Image Segmentation Based on Level-Set Method with Pixel-Region-Dissimilarity Functional

Abstract

Image segmentation plays an essential role in medical image processing There are several methods to accomplish segmentation: one is proposed by Mumford and Shah that utilized the smoother and boundary detector another was proposed by Chan and Vese that took level set to separate different regions In this dissertation my professor and I have proposed a way to distinguish the area of interested by minimizing the dissimilarity in each region Moreover we put a smoother into the functional to reduce the noise The numerical results show that ours can characterize the boundary better than Chan and Vese method In summary the advantages of our approach are robust against noise and give a more clear edge Besides it can not only apply to medical images but the general images adequately

Date of Award	2019
Original language	English
Supervisor	Yu-Chen Shu (Supervisor)