Image Denoising Research Based On Multi-Scale Transform

Images achieved in the real world often are noised. These noise deteriorate images quality and make them blur and even flush their characters and cause difficulty to image apply, analysis. The purpose of image denoising is reduce and remove noise in image and improve image quality. These years, along with the development of wavelet transform theory, and appearances of Curvelet transform, Bandlet transform and Contourlet transform, has make up a rising subject: multiscale analysis, and we call these transform multiscale analysis tools.These tools can depict different frequencies, and some tools developed in these years has overcome the drawback of wavelet. The appearance of multiscale analysis tools bring a reform to digital graphic processing technique, so does it bring a lot to image denoising. In this paper, we firstly summarize the development of multi-scale transform in these years, and compare their characters and differences, and then introduce Wavelet transform and Contourlet transform on image denoising; secondly, analysis the distribute character of noise in wavelet domain, and we proposed two wavelet image denoising method: (1). a image denoising algorithm based on inter correlation of wavelet coefficients at finer scales, the method base on statistics about noise in inter scale relativity, and propose a zero tree like structure, and we combine it with classical Bayes threshold denoising, then make up a novel method, and this method has a better denoising result. (2). Propose a method based image detection and zero-like tree structure denoising. This method absorb the thought of edge detection considering inter-scale dependency, and mend image edge information,. At last, draw in non-local means method, and put it in Contourlet domain。...