Integrated segmentation of noisy image based on the spatial relationship

Abstract

In this paper, we propose a new algorithm for an integrated image segmentation based on the combination of both Markov Random Fields (MRF) and Graph Cuts (GC). In the well-known GrabCut method, the T-link weights do not take into account the spatial relationship between the neighboring pixels. The proposed algorithm, unlike GrabCut method, incorporates this spatial relationship right into the T-link weights. The performance results obtained using natural images clearly demonstrate the robustness, accuracy and effectiveness of the proposed algorithm, as compared to other known methods.