A study on an algorithm of image segmentation based on normalized cuts

Abstract

In this thesis, we investigate a segmentation algorithm based on the normalized cut framework proposed by Shi and Malik (2000). It is a recent approach for solving the perceptual grouping problem in vision. Rather than focusing on local features and their consistencies in the image data, this approach aims at extracting the global impression of an image. We treat image segmentation as a graph partitioning problem and consider a global criterion, the normalized cut, for segmenting the graph. The normalized cut criterion measures both the total dissimilarity between the different groups as well as the total similarity within the groups. We show that an efficient computational technique based on a generalized eigenvalue problem can be used to optimize this criterion. In addition, we suggest a model of similarity function between two nodes and the way to determine some parameters inside such a function.