Unsupervised Texture Segmentation Based on Watershed and a Novel Artificial Immune Antibody Competitive Network

Abstract

This paper presents a novel texture segmentation scheme based on two techniques: watershed and a novel structural adaptation artificial immune antibody competitive network (SAIANet). The proposed scheme first partitions image into a set of regions by watershed algorithm and then clusters the watershed regions by SAIANet, where the gray level co-occurrence matrix and the wavelet frame texture features are extracted from each watershed region as the antigens of SAIANet. A new immune antibody neighborhood and an adaptive learning coefficient are presented, and inspired by the long-term memory in cerebral cortices, a long-term memory coefficient is introduced into the network. The minimal spanning tree in graph theory is used to automatically cluster antibody obtained in the output space without a predefined number of clustering. Finally, the presented SAIANet is devoted to performing a fully unsupervised texture segmentation with a superior performance, which makes full use of the watershed segmentation results.