Context adaptive image denoising through modeling of curvelet domain statistics

bstract. We perform a statistical analysis of curvelet coefficients, istinguishing between two classes of coefficients: those that conain a significant noise-free component, which we call the “signal of nterest,” and those that do not. By investigating the marginal statisics, we develop a prior model for curvelet coefficients. The analysis f the joint intraand inter-band statistics enables us to develop an ppropriate local spatial activity indicator for curvelets. Finally, ased on our findings, we present a novel denoising method, inpired by a recent wavelet domain method called ProbShrink. The ew method outperforms its wavelet-based counterpart and prouces results that are close to those of state-of-the-art denoisers. 2008 SPIE and IS&T. DOI: 10.1117/1.2987723