Image restoration with noise suppression using a multiresolution support.

Abstract

. In Starck & Murtagh, 1994 (SM94), it was shown how noise suppression could be built into widely-used image restoration methods, such as the Richardson-Lucy method. Arising from this work, two issues are resolved in the present paper. Firstly SM94 suppressed noise, based on the supposition that the input image could be considered as a realization of a Gaussian distribution. In this paper, noise suppression using a model based on Poisson noise and additive Gaussian read-out noise is described. Secondly, SM94 found problems in regard to photometric accuracy when suppressing noise. A novel multiresolution method is described here which avoids this problem. The new version of the algorithm is applied to simulated images of point sources and an elliptical galaxy. In both cases the results obtained are compared to the "truth" images, i.e. the unblurred, noise-free images from which the degraded input images were constructed. Key words: Multiresolution analysis -- wavelet transform -- varian...