Keywords: Multiresolution analysis, wavelet, image processing, image restoration, deconvolution
This paper shows how an effective noise suppression strategy can be incorporated into algorithms for the solution of the inverse problem. The residual in the fit of the restored image, at each iteration, is analyzed using a wavelet transform. In order to suppress noise, only significant wavelet coefficients are retained. The effectiveness of this procedure is demonstrated.
Full paper available in HTML and compressed PostScript