| Pulse Coupled Neural Network (PCNN), which shows phenomena of pulse emission, double-channel-multiplying modulation and dynamic threshold variation, is a new generation artificial neural network different from traditional artificial neural networks. It can be applied in every aspects of image processing. Now more and more researchers pay attention to it. At present, it is not long before the research of PCNN has begun, and more work need to do further. Doing research on PCNN has the important significance and value in theories and applications.The thesis chooses a reasonable simplified model by analyzing and comparing. Using it in image smoothing, image segmentation and edge detection, the thesis analyzes and studies the parameters in the model to solve the problem of optimum parameters so that the understanding of PCNN is expanded further.Three new main contributions of this thesis are as follows.(1) In accordance with the pictures which are polluted by two different types of noises---salt-and-pepper noise and Gaussian white noise, the thesis analyzes the parameters in PCNN algorithm one by one in detail, and reaches some serviceable conclusions, pointing out there exist some differences in choosing parameters of the model when removing the salt-and-pepper noise and Gaussian white noise.(2) When studying image segmentation, the thesis analyzes the optimum of the parameters ---the linking strength βij and the initial value of the threshold θij in detail, in the algorithm based on image's entropy and PCNN. Introducing the ultimate measurement accuracy RUMAf as the standard of the degree of the excellence of the parameters, the thesis obtains the principle to select βij. By simulating, it gets thescope of ij.(3) In the research of image edge detection, the thesis analyzes and discusses the parameters of the algorithm, linking with the dynamics equations, and obtains the scope of every parameter, so that the understanding of the algorithm is expanded further.
|
PDF Full Text Request