Wavelet-based color image denoising

This paper introduces a new approach to color image denoising that exploits scale-specific, cross-component edge correspondences inherent in color image data. Processing takes place in the wavelet transform modulus maxima domain wherein image data is represented in terms of a set of multiscale edgemaps. Denoising is accomplished by building noise suppression masks which discriminate noise related edges from physically significant edges. Experimental results show that the new approach outperforms wavelet thresholding in images containing irregular textures.