Image segmentation by deep community detection approach

Abstract

To address the problem of segmenting an image into homogeneous communities this paper proposes an efficient algorithm to detect deep communities in the image by maximizing at each stage a new centrality measure, called the local Fiedler vector centrality (LFVC). This measure is associated with the sensitivity of algebraic connectivity to node removals. We show that a greedy node removal strategy, based on iterative maximization of LFVC, has bounded performance loss relative to the optimal, but intractable, combinatorial batch removal strategy. A remarkable feature of this method is the ability to segments the image automatically into homogeneous regions by maximizing the LFVC value in the constructed network from the image. The performance of the proposed algorithm is evaluated on Berkeley Segmentation Database and compared with some well-known methods. Experiments show that the greedy LFVC strategy can efficiently extract deep communities from the image and can achieve much better segmentation results compared to the other known algorithms in terms of qualitative and quantitative accuracy.