Multiscale Distilled Sensing: Astronomical source detection in long wavelength images

The increasing number of astronomical surveys in mid- and far-infrared, as well as in submillimetre and radio wavelengths, brings more difficulties to the already challenging task of detecting sources in an automatic way. These specific images are characterized by a more complex background than in shorter wavelengths, with a higher level of noise, more noticeable flux variations and both unresolved and extended sources with a higher dynamic range. In order to improve the source detection efficiency in long wavelength images, in this paper we present a new approach based on the combined use of a multiscale decomposition and a recently developed method called Distilled Sensing. Its application minimizes the impact of the contaminants from the background, unveiling and highlighting the sources at the same time. The experimental results achieved using infrared and radio images illustrate the good performance of the approach, identifying greater percentages of true sources than using both the widely used SExtractor algorithm and the Distilled Sensing method alone.