Image Denoising Method Based On Wavelet And NSCT

The image we get in our life has often been affected by the noise. The existence of noise makes it difficult for the subsequent processing and analysis of image. Therefore, it is very meaningful for researching image denoising. In this paper, we mainly research the image denoising based on wavelet transform and NSCT transform, the mainly tasks are as follows:(1) An improved NLM algorithm based on wavelet domain is proposed. The method is based on existing NLM method in wavelet domain. In order to estimate the similarity between the image pixels more accurately, this paper considers the similarity of low frequency image and high frequency image at the same time, and we use the product of low frequency image similarity and high frequency image similarity as the final similarity between the image blocks of every details bands. Then weight value calculation formula of improved NLM method based on wavelet domain is proposed. The experiment results show that the improved algorithm can get better effects in noise suppression and preserving image edges and details compared to the original algorithm.(2) For ultrasound image despeckling, the bivariate threshold denoising method based on orthogonal wavelet transform is expanded to Daydic wavelet transform domain. A bivariate shrink despeckling method for ultrasound image based on Daydic wavelet is proposed. The translation invariance of Daydic wavelet and good response characteristics for edges and details makes the method can better preserve the details of the ultrasound image.(3) For SAR image despeckling, the despeckling method for SAR image based on NSCT is proposed. Bringing in non-logarithmic additive noise model, the distribution characteristic of non-logarithmic additive noise in NSCT domain, and the Guassian model is used to model the non-logarithmic additive noise of the local isotropic region in NSCT domain. Then, the high frequency coefficients are shrinked by SURE-LET. The experimental results show that the proposed algorithm effectively smooths the speckle of isotropic region, meanwhile, it can better preserve the radiation characteristics and the structure information of SAR images.