The Image Processing Based On Pulse Coupled Neural Networks

Image processing is the most important step during the analysis of image. It turns origin image into abstract data which is used for analyzing and understanding image. Although it was been developed and focused on from 1970, the general method and the objective standard, which is used to judge the result of image processing, have not been found.Neuron network is one of the most important methods for the image processing. In 1980's early, it was used in recovering image. Since it is based on biologic vision system, it has a lot of advantage which traditional methods never have. PCNN (Pulse coupling neuron network) is the best one of neuron networks.In this article, the research on PCNN includes setting up mathematic model for PCNN and validate this model by apply it to image process.First, reviewing PCNN model and the development of image processing, analyzing pulse mechanism and working principle of PCNN, explaining how to predict the result of PCNN, when the number of neuron network and linking between neurons are few.Second, introducing PCNN to color image segmentation, choosing fit color space, avoiding relations between components, applying PCNN to every component, and then combining the results. As it is known, it is the first time that PCNN is applied to multidimensional.Third, reviewing the relation between mathematical morphology and PCNN to set up mathematical model for PCNN. Using set theory to establish model for PCNN based on the relation. And then utilizing this model to analyze the previous PCNN method for color image segmentation, obtaining the advantage and disadvantage. Then, making use of vector ordering based on set theory, designing new method for color image segmentation. The result is the same to the one from model. The model is enhanced.Finally, the problems requiring further studies are discussed.