Wavelet Transform In Image Noise Suppression

With the popularization of the internet and the development of the multimedia technology, images become one of the most important ways to acquire information. However, images are easy to be polluted by all kinds of noise in its acquisition or transmission, which degrades the quality of original images and has unfavorable influence on consequent processing. So it is necessary to suppress image noise. The purpose of noise suppression is to highlight useful information, weaken or remove some disturbances so as to obtain more suitable data for human or machine analysis and processing.Currently, a large number of noise suppression approaches are proposed. Most traditional methods do not take it into full consideration that how to maintain useful information such as image edges or image texture, so the images may become blurred or some details are lost in the process of noise suppression. In recent years, wavelet transform has been widely used in noise reduction for its well time-frequency analysis features. This dissertation researches on the application of wavelet transform to reduce image noise, whose main contribution are as follows:(1) After the NormalShrink method suggested by Donoho's VisuShrink and Lakhwinder Kaur is analyzed, an improved adaptive wavelet threshold algorithm on noise suppression is proposed. According to the different characteristics of noise at different wavelet scales and the different noise intensity, an optimal threshold of each scale is determined via the estimation of noise variance. Experimental results show that our algorithm is superior to some traditional spatial filters and new-developed methods based on wavelet both in visual effect and peak signal noise ratio (PSNR).(2) To improve heavy noise-polluted images quality, a new image enhancement algorithm is proposed. Firstly, the noised image is transformed into wavelet domain via multiscale decomposition, and the information of different resolutions and different directions in frequency domain are obtained. Then the regularities of coefficient distribution and attenuation characteristics of noise and edges in different subbands are employed to distinguish noise from edges and textures, thus the image quality is improved by noise suppression and edge enhancement. Experimental results show that our algorithm is superior to most;traditional spatial filters and new-developed methods basing on wavelet or contourlet transform both in visual effect and Peak-Signal-to-Noise Ratio. It can be applied to image enhancement preprocessing in heavy noise environment.