摘要
A new locally adaptive image denoising method, which exploits the intra-scale and inter-scale dependency in the dual-tree complex wavelet domain, is presented. Firstly, a recently emerged bivariate shrinkage rule is extended to a complex coefficient and its neighborhood, the corresponding nonlinear threshold functions axe derived from the models using Bayesian estimation theory. Secondly, an adaptive weight, which is able to capture the inter-scale dependency of the complex wavelet coefficients, is combined to the obtained bishrink threshold. The experimental results demonstrate an improved denoising performance over related earlier techniques both in peak signal-to-noise ratio （PSNR） and visual effect.
A new locally adaptive image denoising method, which exploits the intra-scale and inter-scale dependency in the dual-tree complex wavelet domain, is presented. Firstly, a recently emerged bivariate shrinkage rule is extended to a complex coefficient and its neighborhood, the corresponding nonlinear threshold functions axe derived from the models using Bayesian estimation theory. Secondly, an adaptive weight, which is able to capture the inter-scale dependency of the complex wavelet coefficients, is combined to the obtained bishrink threshold. The experimental results demonstrate an improved denoising performance over related earlier techniques both in peak signal-to-noise ratio （PSNR） and visual effect.
基金
This work was supported by the National Natural Science Foundation of China under Grant No. 60573027.
