Publication Details
Linda Tessens, Alexandra Pizurica, Alin Alecu, Adrian Munteanu, Wilfried Philips

ProRISC 2006, Veldhoven, The Netherlands

Contribution To Book Anthology


In this paper, we perform an inter-sub-band statistical analysis of curvelet coefficients, making a distinction between two classes of coefficients: those representing useful image content and those dominated by noise. This analysis enables us to develop an appropriate inter-sub-band local spatial activity indicator (LSAI) for curvelets. We use this LSAI in our recently developed curvelet-based denoising method ProbShrinkCurv. The results demonstrate that the new method outperforms the wavelet-based ProbShrink estimator as well as the existing curvelet-based methods, both for textured and for piecewise smooth images.