| Evolving from the research on the visual cortex of small mammals likecats, the Eckhorn model[1] is the basis of Pulse Coupled NeuralNetworks(PCNN). Johnson and many other researchers put PCNN to useafter further modifications. PCNN is different from those models oftraditional artificial neural network. It is constructed by simulating theactivities of the neurons of visual cortex, and is the simplification andapproximation of the neurons. The properties inherited in PCNN, such aslinking field and the dynamic threshold, make the approximate neurons firesimultaneously, and this is very close to the natures of visual cortex of smallmammals. So it has gained widely applications in image processing, such asimage segmentation, edge extraction, object recognition and so on.At the same time, PCNN is a neural network with many parameters, andthe performance of PCNN highly depends on the set of these parameters. Sofar, the optimal parameters of PCNN for different images can only bequalitatively analyzed and there is no way to get the optimal valuesautomatically. So the adjustments of these parameters become a key andonerous work. In this paper, through reducing some parameters we get asimplified PCNN which also has the linking field and the dynamic threshold.Then the simplified network model is applied to image denoising andsegmented image coding, and its performance is analyzed and summarized.In Chapter two, the initial research on the removing of Impulse noiseand Gaussian noise is done. On the basis of the predecessors'work andcombined the local median algorithm, the noise filter based on simplifiedPCNN is presented. The experiment results are compared with otheralgorithms and at last analyzed and concluded.Chapter three analyzes applications of PCNN on image segmentation,uses the simplified PCNN as the segmentation algorithm of segmented imagecoding, and brings forward segmented image coding based on PCNN. Andthe experiment results are compared with traditional block transform coding.
|
PDF Full Text Request