Research On Image Denoising Algorithm Based On Wavelet Transform And Support Vector Machine

The information conveyed by the image is direct-viewing, vivid and easily understoodand received,in real life, people get most of the information provided by the image. However,in recent years, with the progress of society and the rapid development of science andtechnology, the image in filming, compression, storage, and in the transmission process,adding a lot of unnecessary noise inevitably for interference by imaging equipment andenvironmental factors，making the most of the image become blurred, lower quality, visualeffect is not optimistic, even there is a serious threat to people’s right to judge and use pictures.In order to allow people to get a clear image, reducing the losses caused by image qualityproblem, denoising is particularly important for noisy image. A good image denoisingalgorithm can not only to remove the noise as much as possible but can not affect theimportant details of the image. Images are the original image and the noise, approximately, wecan consider image denoising to be a classification process. From the perspective ofclassification the nature of image denoising is the process of separating the noise and theoriginal image.SVM in classification problem can be used widely, because it is a typical classification, ithas many advantages, can be widely used in the classification problem for dealing with highdimension, small sample and generalization performance. Based on the properties of waveletcoefficients and the advantages of SVM, in this paper, a wavelet-based image denoising usingSVM the nature of the wavelet coefficients is proposed. In terms of feature vector selection,noise and the characteristics of the wavelet coefficients of the original image will bestrengthened by a coefficient algorithm, and then selecting noise coefficient and thecoefficient of the original image as the feature vector machine training input, and thecorresponding coefficient of weight will be treated as the output of the training machine. Thento deal with noise pixels in the image,it is divided into two different types: noise point andnon-point noise by support vector classifier and thresholding it. Finally, using the Matlab to simulate the proposed algorithm.Simulation results show that the method has a gooddenoising effect, can achieve higher peak signal to noise ratio.