Binary clustering of color images by fuzzy co-clustering with non-extensive entropy regularization

This paper proposes semantically meaningful binary clustering of color images by a novel fuzzy co-clustering algorithm. The clustering objective function incorporates the non-extensive entropy with Gaussian gain for regularization purpose. The chromatic color components in the CIEL∗A∗B∗ color space form the feature space for clustering. The result is a very good differentiation of the colors in the scene as belonging to the foreground object and the background, which helps in scene understanding and information gathering. One direct application of our tool is salient or foreground object segmentation. Experimentation on images from a benchmark dataset and comparisons with the state of the art clustering and segmentation methods establish the efficiency of our approach.