Research On Image Denoising Based On Wavelet Transform

Image denoising is one of the most diverse areas of research in image processing and computer vision. We all hope that after the image denoising the picture could be able to have a good visual effect and high precision, and can keep the important features of image, such as edge, texture and sharp part. Based on this purpose, this thesis proposes a new threshold function for wavelet image denoising. In addition, the value of new threshold function is optimized to improve the efficiency and effectiveness. The threshold is optimized in two cases, supervised learning and unsupervised learning. At the same time, for different characteristics additive Gaussian noise and multiplicative speckle noise,different threshold training methods are used. The proposed method can be used to suppress the noise of the two main classes: additive gaussian noise and multiplicative speckle noise.Denoising technology can effectively suppress noise, but it may cause the image edge blur, which affects the quality of image denoising. In order to get a better image visual effect, the techniques of edge detection is used to extract the denoising image edge information. Put the edge information at the corresponding location of the image of denoised,to restore some denoising mistaken for noise edge points, to improve the image details.Simulation results show that the improved threshold function has relatively good denoising effect compared to the classical soft thresholding function and hard threshold function and some commonly used threshold function. Optimized threshold either under the supervision of learning or unsupervised learning, both have better results compared to other thresholding method(Global threshold, Miniman threshold, Bayes threshold). And the image after edge extraction has a slightly higher peak signal to noise ratio.