Abstract: This paper proposes a fully unsupervised image segmentation algorithm by using a novel structural adaptation artificial immune antibody competitive network without a predefined number of clustering. Based on the basic conception of self organizing feature map, a new immune antibody neighborhood is presented to enhance the robustness of the network, and inspired by the long-term memory in cerebral cortices, a long-term memory coefficient is introduced into the network to improve the convergence speed of the algorithm, and three death operations are presented to eliminate those antibody droves by noise antigen. With above advanced methods, the model can adaptively map input data into the antibody output space, which has a better adaptive net structure. This approach is applied to segment a variety of images into homogeneous regions, including synthetic texture images, remote images and SAR images, and experimental results illustrate the effectiveness of the proposed novel algorithm.

Key words: unsupervised image segmentation, artificial immune networks, structural adaptation, data clustering