Image segmentation based on community detection approach

Image segmentation is a fundamental low-level vision problem with a great potential when it comes to its applications. Several methods exist to split an image into regions. However, this problematic is still a relatively open topic for which various research works are regularly presented. With the recent developments in complex networks theory, methods based on graphs, which, can segment an image has considerably improved. This paper presents a new perspective of image segmentation by applying the most efficient community detection algorithms. For this, we first transform images into an adjacency graph. Then, we propose to study five complex network dedicated community detection methods which are Infomap, Louvain, Fast multi-scale detection of communities based on local criteria, Multi-scale detection of communities using stability optimization and stability optimization based on Louvain. Finally, we extract communities (regions) in which the highest modularity or stability feature is achieved. In our experiments, we establish a fair comparison between the proposed algorithms for Berkeley database images, and we show that a good performance is achieved by multi-scale detection of communities using stability optimization with a probabilistic rand index PRI of 0.81.