Research On Image Denoising Based In The Transform Domain

The higher image quality requirement is needed with the development of informational network. But in fact, the images inevitably will be affected by noise interference in the process of transmission and processing. Noise reduces the image quality, makes the image become fuzzy, and even covers the image characteristics. Therefore, Image denoising is always an important issue. How to remove the noise and preserve fine image details is also an important topic.In order to remove the noise and improve image quality, this thesis further study image denoising technology from the angle of the transform domain denoising and reserve the detail information. First of all, a neighboring adaptive bayesian shrinkage image denoising method in dual-tree complex wavelet domain is proposed in view of the superiority of wavelet processing in the image denoising. Secondly, a block matching and3D filtering algorithm is optimized through sharpening image details in order to enhance image clarity.In order to remove noise which is introduced by image acquisition or transmission more effectively, the neighboring adaptive bayesian shrinkage image denoising method in dual-tree complex wavelet domain made use of the translation invariance and the advantage of more direction selective of the dual-tree complex wavelet transform, and the local adaptive neighborhood correlation of the coefficient was also considered. The variance of the corresponding coefficient of the appropriate neighborhood full inch window was estimated, the average of the variance which was used as the variance of the whole sub-band image was calculated using the sliding window. Bayesian shrinkage method was used to handle the wavelet coefficients to achieve efficient denoising. The experimental results show that the proposed method gets higher PSNR and better visual expression. The denoising performance is excellent.In order to further attenuate noise while sharpening image details, this chapter proposes a block matching and3D filtering algorithm based on alpha-rooting. The algorithm sharpens image by alpha-rooting, enhancing fine image details and reflecting image texture features obviously. Then the algorithm further removes noise by the block matching and3D filtering algorithm. The experimental results show that the improved algorithm can preserve and enhance fine image details and effectively attenuate noise.