| With the introduction and application of wavelet transform in the processing of picture signal , it make digital picture processing to a higher extent. Wavelet analysis have many advantages over discrete fourier transform by zoom-in and zoom-out capability .Wavelet denoising in multi resolution approximation has been evolved to become very powerful tools for digital picture signal processing. Hence Many more people has been paied attention to the wavelet transform, because wavelet analysis plays a leading part role in the digital picture processing. .As far as we are concerned, up to the present time ,paper from home and abroad mainly focused on two problem of wavelet analysis in the process of picture processing .one is wavelet denoising the other is picture compressed, in term of the level of wavelet soft threshold of Donoho , John stone , Panquan and Meiwenbo. Based on the least-mean-square algorithm But there is a deficiency in the determination of wavelet threshold of the different resolution, generally speaking, the wavelet soft threshold is determined by noise statistics characteristic, which is drawn by Autoregressive modern spectrum estimates in the presence of noise. It is well known to us all. picture is polluted by complicated pulse noise. Noise statistics characteristic estimation does not live up to the best standard. A new resolution presented in this paper to this problem gives rise to increase the ratio of signal-noise by suggesting I set up a relation between the ratio of signal-noise and wavelet soft threshold . Only by mean of the resolution , Can we select the signal that is the higher- ratio of signal-noise And cut off the lower ratio of signal-noise, thus the globe ratio of signal-noise is enhanced to a higher degree by the enhancement of the local ratios of signal-noise. There are three main reasons attributed to this result as follows:Firstly, I choose a wavelet transform named fivezhu wavelet <, Which is different the other wavelets from the fact that fivezghu wavelet have seven7sample points and was decomposed a orthogonal, two-band filter bank and include a detailed part; which can be realized by two orthogonal mirror symmetries lattice filter. Because orthogonal wavelet packet makes sure of not only tight frame , orthogonal, symmetries , and high order vanished-moment but linear phase as well. And its resolution is two times of the that of the two order wavelet. More often than not. Usually, we consider a sampling technique in which each sample is obtained by averaging a diamond portion of the picture, as we know, a large amount of information of a image in nature mainly gathered the small area, at the same time <, .but to finger print gray image , A fine degree of quantization is particularly important for samples taken in regions of a picture across which the gray level changes slowly. Secondly: I choose a non-linear filter named hybrid filter which combines linear and nonlinear filters is proposed for print gray image, it Performs better than average filters and median filters on noise reduction .While retaining edges of an image. I suggest a modified version and fast implementation of this filter improve its performance. Because of its reduced computation complexity, this filter excels in real-time tasks. Thirdly: during the course of gray image processing, we make gray image through filter, then give output of the filter wavelet transform. Output can be obtain, next we use our designed the wavelet soft threshold to select result of the wavelet transform , finally , we give the selected result reversal wavelet transform .It is obvious: the wavelet soft threshold is important to improve the quality of the gray image processing.. I give the Donoho wavelet soft threshold a modified value method , which has a relation with ratio of signal-noise. I made full use of discrete Hop field single feedback neural network , and nonlinear steady of automatic system at last , I obtained a steady limited ring, give the energy function an order differential a...
|
PDF Full Text Request