Multiresolution image reconstruction by wavelet decomposition

We present a method of image reconstruction based on wavelet decomposition. The method seeks to overcome the problem of noise amplification during the image reconstruction process. We decompose the raw image to be reconstructed into several wavelet planes and a residual image, we reconstruct each one independently using an iterative maximum likelihood algorithm. To control the process, we stop the reconstruction of each one of the wavelet planes and the residual image at a different number of iterations. The stopping points are determined using the cross-validation method. The method has been applied to data from the non-refurbished Hubble Space Telescope.