A New Approach for Wavelet Denoising Based on Training
Zelong Wang, Fengxia Yan, Jiying Liu, Jubo Zhu · 2009
A new approach for wavelet denoising based on training is proposed in this paper. Firstly, the same two images with different noise levels are trained for the parameter of SLT (Slicing the Transform). Secondly, SLT is used to remove the noise iteratively. The benefit of conventional wavelet denoising, such as multi-analysis, is reserved by this paper. Furthermore, the difficulty of selection about natural images for training is avoided and the solidity of the algorithm is enhanced as well as the speed. Experimental results show that the proposed method is effective to a wide range of images; when compared to the classical method, the reconstruct images with our proposed method are with better PSNR (peak signal-to-noise rate) and visual quality.