ADAPTIVE IMAGE SEGMENTATION BASED ON SELE CTIVE MULTIRESOLUTIONAL KOHONEN NEURAL-NETWORK

As to the widely used multi-thresholding technique, a new clustering m ethod was proposed, namely the Adaptive Image Segmentation Technique B ased on Selective Multiresolution Kohonen Neural Network. The proposed approach has the advantage that the number oi segmentation scales is not needed to specify a priori. It can be adjusted dynamically accordi ng to local statistic property of the image.