A Method Of Image Hybrid Noise Filtering Based On Simplified PCNN

Pulse Coupled Neural Network (Simplified as PCNN) is a new kind of network proposed resently which has biological background, it is the simplification and approximation of the real neurons which is different from the traditional artificial neural network. The PCNN model has the link and dynamic threshold fading character which make the approximate neurons fire simultaneously, it can make up the space incoherence of data-in and small change of the amplitude and can retain the image region information completely, it has been used in Image Segmentation, Image Filter, Target Recognition, Feature Extraction and so on.This thesis firstly illustrate the basic theory and principle briefly of Pulse Coupled Neural Network, analyzed its structure character and basic speciality, simplified PCNN model appropriately on the foundation of refering existent theory and ideas, simplified the structure of original model, reduced the parameters of the model, reserved several important specialities of the original model.Secondly, this thesis analyzed the speciality of mixed noise constituted by pulse noise and Gauss noise, applied the simplified PCNN model to image mixed-noise filter, utilized several technique specialities of the model, selected parameters properly, eliminated noised and reserved detail imformation of image at the same time.Through simulation experiment,we can confired that the effect of eliminating gray image mixed noise which applying the simplified PCNN eliminating algorithm proposed in this thesis is good. And introduced two traditional noise eliminating algorithms, comparation proved the simplified PCNN eliminating algorithm is better the traditional noise eliminating algorithms. With the increasing of the image pouluted by blends noise, the advantage is obvious. After carrying out corresponding colour space change aimed at the colour image, built a filtering algorithm based on the blends noise in colour image of the simplified PCNN model. Through lots of experiments, we can confirm that the filtering algotithm of simplified PCNN model also has a good effect in the field of filtering blends noise.