Research Of Image Segmentation Based On Pulse-coupled Neural Network

Pulse Coupled Neural Network (PCNN) is constructed by simulating the activities of the mammal's visual cortex neurons. This model has the characteristics of grouping the pixels which have similar two-dimensional space and gray level, and can diminish the partial gray differences, repair the local and minor interruptions of the image. It is a method of image segmentation that conforms to human vision characteristics.Fisrt of all, according to the property of biology neuron of the ionic mechanism of postsynaptic inhibition and based on traditional PCNN, the model of ignition from bottom to top is provided.Sencondly, when a image is divided witn traditional PCNN, the pixels and neurons are one to one, which makes the velocity of network slow. To solve the problem, Bidirectional search Pulse Coupled Neural Network (BPCNN) is provided in this article. Different from the former PCNN, the model conforms to the mechanism of biology neural network and has two size thresholds, of exciting type and of depressing type, which cause the network to ignite from top to bottom and from bottom to top at the same time .It can search the best threshold from two directions together with max entropy. The BPCNN can speed up network operations. The validity of this model is well verified by experiments of grey images.Finally, compared with grey images, color images supply brightness, hue , saturation and more information .How to make best use of the information is the key and difficult point. To solve the problem of color segmentation, this article combines BPCNN with Triggered Pulse Coupled Neural Network based on receptive field (RTPCNN) , and creates Double-PCNN (DPCNN) based on BPCNN and RTPCNN. The new model mixes their merits that speed up operations and enhance image segmentation .The validity of this method is well verified by the analysis and experiments .The result of segmentation is well suited for further recognition.