The Research On Image Denoising Algorithm Based On Wavelet And NLM

The development of computer technology and mathematics provides a strong foundation for digital image processing technology. With the enriching of material life and spiritual life, people don't only passively accept more and more visual information. but also tend to perfectly process the visual information using digital image processing technology in advance, to meet the higher requirements finally.Because of various factors, the process of image processing is inevitably interfered by a lot of noise. which causes substantial decline of image information quality, and brings difficulty to image analysis at the same time. So it is very necessary to do image pre-denoising in advance of analyzing image information.There are two improved image denoising algorithms in this paper:improved image denoising algorithm based on wavelet and improved image denoising algorithm based on NLM.In the past few years, many researchers in and aboard devoted to wavelet fields. They have founded a complete theoretical system, and made outstanding contributions to image denoising. However, because wavelet coefficient of high-frequency component includes edge and noise information, traditional wavelet transform can not both denoise and keep the image information (such as edge information) at the same time. An improved denoising algorithm based on wavelet is proposed in this paper:Firstly, an improved edge detection method based on wavelet spatial correlation is put forward. then Zerotree-Like Structure bayesian threshold method is used for denoising, finally the extracted edge information is replenished where the edge information is lost during the Zerotree-Like Structure bayesian threshold denoising.NLM image donoising algorithm proposed by Buades introduces new thoughts for image donoising fields. Practice shows that NLM has lots of advantages, but it can not adaptively adjust effectively to images with different content or with different structures. There are two improved algorithms based on NLM in this paper. (1) Firstly based on NLM. do statistics of different structures of image according to wavelet energy. Then obtain the energy-parametric funtion through Binary Fitting, and use the function to get the optimal MLM denoising parameter. Finally denoise the whole image using NLM and achieve the best effect of image denoising. (2) Firstly segment the input image with Mean Shift algorithm. Then carry out self-adapting NLM image denoising towards each segmented image block. Result shows that the algorithm has steady and efficient result.