The Research Of Image Enhancement Technology Based On Pulse Coupled Neural Network And Fuzzy Theory

Image enhancement technology is an important part of image processing technology whose purpose is processing the images in order to obtain the better visual effect and more useful images according to special applications, so that it lays the foundation for the following image analysis, understanding and recognition. Conventional image enhancement technologies are simply changing the contrast of images or controlling the noises, at the same time the details of the images are weakened, user interventions are more needed and can't enhance the images automatically and so on.Aimed at the defaults of traditional image enhancement algorithms and the needs of image enhancement in special enviroments, a new fuzzy enhancement algorithm based on improved pulse-coupled neural network and improved genetic algorithm is proposed. Pulse coupled neural network (PCNN for short) is a new type of neural network recently proposed. It was established on the basis of Eckhorn Model in which nerve cell activities in biology visual layer were simulated. Johnson and other researchers have done more research about the model which finally evolved to be a monolayer iterative network model and it much more accorded with biologic natural visual characteristic. The model was abroad applied in various fields of image processing such as removing noises, image segmentation and so on.The work of this paper mainly consists of three parts: Firstly, with deep research on the work mechanism of pulse coupled neural network the model will be improved and its important character is adequately utilized: neurons joined to each other exists the energy spread which can make the similar neurons output synchronized pulses.Emphasis study is its application on image smoothing or removing noises. Secondly because edges of the image needed to enhance have uncertainty which is fuzzy character so it is better to apply fuzzy theory into image enhancement than traditional enhancement algorithms. It is much more according with practical situations and can obtain better enhancement results. This paper aims at the defaults of exist algorithms such as choosing threshold at random, the ability of noise cancelling is weak and so on. Two kinds of improved simple level and multi-levels image enhancement algorithms based on fuzzy theories are proposed. Thirdly because the idea of"first segmentation next enhancement"was adopted in the paper and confirmation of the transition point comes down to selection of best segment threshold value of the image, so the conventional genetic algorithm is modified in this paper. Simple crossover operator and dynamic mating probability are proposed. Simple crossover operator can reduce the possibility of appearance of repeat bunches in the population. Dynamic mating probability is used to avoid prematurity in GA.In this paper pulse coupled neural networks are combined with other algorithms and some new algorithms are proposed by making good use of PCNN itself and the advantages of other efficient algorithms. Some simulation experiments are carried out on software platform of MATLAB 7.0 and experiment results are analyzed and summarized at last.