Wavelet-based denoising considering interscale and intrascale dependences

Recently, wavelet analysis has been used to denoise a digital image corrupted by noise in the acquisition step. Because of the multiscale decompositions (multiresolution) and translation parameter (locality in space), in addition to the scale parameter of the wavelet there are two types of correlations that can be used to detect edges. Many previous works have exploited the intrascale or interscale dependences alone to denoise the image. In this paper, a denoising method that combines the intrascale and interscale correlations simultaneously is proposed. By manipulation of the wavelet coefficients in successive bands, this noise model is investigated exhaustively and estimated as the well-known Gaussian distribution. With the investigated noise distribution, a new denoising method is proposed. Experimental results show the superiority of the proposed method to the interscale or intrascale method according to objective and subjective criteria.