Figure-ground Separation by Contour Statistics and Markov Random Field Model

Abstract

In this paper,we propose a Markov random field(MRF) based representation for the Gestalt law,and suggest using a message passing-like scheme to infer the segmentation.Different from other grabcut models,our MRF function is specially encoded to consider orientations along the contour,thus the Gestalt law is embedded into the inference. As a basic framework of the research in figure-ground separation and Gestalt law,our system is designed in reference to neurophysiology,and the architecture is composed of three modules:primal visual cortex(V1),extra-striate cortex (V2),and the interest selected region.To validate our method,we conduct experiments in both auto and interactive segmentation algorithm.The results are better than those of grabcut and other related algorithms.