Deconvolution

Standard CLEAN algorithms must be slightly modified to work on a dirty mosaics. Indeed, the use of truncated primary beam in the above equations is only a first order measure to avoid noise artifacts. However, the noise level still increases at the edges of the mosaic, implying that at some point the CLEAN algorithms will confuse noise peaks at the mosaic edges with true signal. To avoid this, the iterative search is made on a signal-to-noise image $M(\alpha,\beta)/\sigma(\alpha,\beta)$ instead of the residual image. At the restoration step, the clean component list is used to produce the residual map and clean map. Nothing particular is done with the remaining signal-to-noise image.