IMAGE SEGMENTATION VIA BRITTLE FRACTURE MECHANICS (MP-P4)

Author(s) : Wei Wang (Dept. of ACAE, Chinese Univ. of HongKong, China) Ronald Chung (Dept. of ACAE, Chinese Univ. of HongKong, China)

Abstract : Brittle material has the nature that it is hard to break, but when it is to break it breaks into large, separated pieces whose boundaries are those that are smooth and with dense concentration of strong stress. Image segmentation desires similar results: partitioning an image into regions that correspond to respective surfaces or objects in the imaged scene, which are often large and whose boundaries are smooth, closed, and with dense concentration of strong intensity gradients. In this work we explore the adaptation of the brittle fracture mechanics toward discontinuity detection in the image segmentation process. An image is regarded as a plate experiencing brittle fracture with external forces exerting at positions where the intensity gradient has a large magnitude. Stress concentration effect is considered in the formation of cracks. Real image results indicate the model indeed could have promising performance.

Menu