The Research And Application Of Wavelet Neural Network On Denoising And Compression Image

Based on the theory of wavelet analysis, wavelet neural network is to establish a tiered, multi-resolution one of a new type of artificial neural networks.when wavelet function replace the traditional neural network activation function,a wavelet neural network is constituted.The theory proved that wavelet neural network has the capacity of consistent approximation and L2 Approximation,To ensure reasonable initialization parameters,it has a very fast convergence rate.that is no doubt that the use of wavelet neural network denoising and compressing image,is a shortcut.In this paper, wavelet neural networks was applied to denoise image, strong classifier, in the noise detection,was structered.the computer simulation also showed the obvious advantage of this method ,this theory ensures that the method is more rational than the traditional median filter.At the same time, based on the traditional image compression in KL transform the question of higher-order matrix could not be dealt with, Extraction of principal components analysis of the hebb algorith caused error by data loss problem during the process of image compression, The wavelet neural network can deal with high-level transformation matrix and can achieve convergence in a short period of time.Therefore,above image compression,the article established the ideology of theory based on wavelet neural network combing KL transform, and computer simulation tests verify the reliability of the algorithm.The full text is divided into six chapters, each chapter are as follows: The first chapter summary the development of the history ,significance of the status quo, the value of the paper, the value of practical application and theoretical significance.The second chapter summaried the basic knowledge of neural network and wavelet transform,as,the wavelet neural network structure was constructed.The third chapter established learning algorithm of wavelet neural network, derived by Wavelet frame, learning algorithm of BP wavelet neural network,Also discussed the issue of parameter initialization, learning rate adjustment.The fourth chapter proved the theory wavelet neural network have the capacity of a consistent approximation and L2 ApproximationThe fifth chapter designed weak and strong classifier of noise detection in accordance with the characteristics of image noise. In theory and practice ,affirms the obvious advantages of. the denoising effectThe sixth chapter studied the algorithm through the combination of wavelet neural network and KL transform in image compression,And the computer simulation showed that,Compared to traditional KL transform and Hebb algorithm in image compression, The new algorithm has great advantages.