Texture Feature Extraction And Texture Segmentation Based On Clustering And Classification With Pulse Coupled Neural Networks

This thesis investigates the model of Pulse Coupled Neural Network (PCNN), which includes its basic principles, behaviors and characteristics. The paper describes the characteristics of grouping by similarity, synchronous pulse bursts and capture of the modified PCNN model, and presents a method of data clustering and pattern classification based on PCNN, which has great adaptability and flexibility. The application of PCNN is applied to the field of Pattern Recognition innovatively. In the paper, by analyzing the features of texture images and discussing the various methods of extracting textural features from images, a statistical feature extraction method based on moment images of the texture image and their nonlinear transformation is presented. Both clustering and classification method based on PCNN are applied in texture segmentation of image, which are successfully applied to the texture image segmentation task. At last, the factors which affect the quality of image segmentation are discussed, followed with an example of textural image segmentation, showing great effectiveness of the method.