Application Of Pulse Couled Neural Network In Image Segmentation

Since1990s Pulse Coupled Neural Network (PCNN) has been a new type of neural network with much research value. This thesis studies the theoretical model of PCNN for a specific application object, improves the traditional PCNN model and uses the proposed model to solve image segmentation.Image segmentation is the key steps of image analysis of, which is basic in computer vision technique. Image segmentation is to separate characteristics of the area and extracts the target of interest then processes, for example, forests, cities, roads and farmland segmented from airborne remote sensing photos.OTSU algorithm makes use of maximum variance to segment the threshold of image pixels. The thesis proposes a new segmentation model based on OTSU threshold and PCNN, which simplifies the traditional PCNN model, reduces the parameter selection; improves the traditional PCNN exponential decay of the threshold. The experimental results show the proposed model is faster thannditional PCNN based segmentation methods because it does not need to recursively determine the threshold.This thesis studies the application of the pulse coupled neural networks in image processing. First the OTSU threshold and PCNN theory is analyzed. Then a improved PCNN segmentation model is proposed based on OSTU threshold, which avoid difficult selection of PCNN parameters. The thesis uses the proposed method to segment images and compare it to the classic image segmentation algorithms and traditional PCNN segmentation model. The experimental results show that the proposed OSTU-PCNN segmentation method is feasible.