Probabilistic winner-take-all segmentation of images with application to ship detection

A recent neural clustering scheme called "probabilistic winner-take-all (PW TA)" is applied to image segmentation, It is demonstrated that PWTA avoids underutilization of clusters by adapting the form of the cluster-conditiona l probability density function as clustering proceeds. A modification to PW TA is introduced so as to explicitly utilize the spatial continuity of imag e regions and thus improve the PWTA segmentation performance. The effective ness of PWTA is then demonstrated through the segmentation of airborne synt hetic aperture radar (SAR) images of ocean surfaces so as to detect ship si gnatures, where an approach is proposed to find a suitable value for the nu mber of clusters required for this application. Results show that PWTA give s high segmentation quality and significantly outperforms four other segmen tation techniques, namely, 1) K-means, 2) maximum likelihood (ML), 3) backp ropagation network (BPN), and 4) histogram thresholding.