Graph Cuts for Image Segmentation

In computer vision, segmentation is the process of partitioning digital image into multiple regions (sets of pixels), according to some homogeneity criterion. The problem of segmentation is a well-studied one in literature and there are a wide variety of approaches that are used. Graph cuts has emerged as a preferred method to solve a class of energy minimization problems such as Image Segmentation in computer vision. Boykov et.al[3] have posed Image Segmentation problem as Energy Minimization in Markov Random Field and found approximately minimum solution using Graph cuts. Min-Cut/Max flow algorithms for Graph cuts include both push-relabel methods as well as augmenting paths methods. Boykov and Kolmogorov [2] have developed an efficient method for finding augmenting path. Though experimental comparison shows this algorithm efficient over other, worst case complexity of it is very high. In [1], Voronoi based Push Preflow method is suggested to find min-cut, which not only exploits the structural properties inherent in image based grid graphs but also combines the basic paradigms of max-flow theory in a novel way. Though Min-cut/Max-Flow based Graph cut methods can efficiently find partitions, those (partitions) may not be the desired ones. So, [7] have developed concept of Normalized cuts. Normalized cuts considers association within a cluster as well as the disassociation among clusters.