JOINT SEGMENTATION AND REGISTRATION VIA VARIATIONAL FORMULATION FOR 2D MONO-MODAL IMAGES

Abstract

Two of the main branches of image processing are segmentation and registration. Image segmentation aims to partition the images into foreground and background based on distinguishable characteristics, meanwhile, image registration involves finding an optimal geometric transformation from the given images. These two tasks are often treated separately, however, the joint models between segmentation and registration have their advantages compared to the separate tasks. In this paper, two variational models for joint image segmentation and registration are reviewed and compared using mono-modal images. Results showed that joint image segmentation and registration model based on linear curvature gives better results compared to the state-of-the-art models.