A procedure is described for estimating an optimum kernel for the detection by convolution of signals among Poissonian noise. The technique is applied to the detection of x-ray point sources in XMM-Newton data, and is shown to yield an improvement in detection sensitivity of up to 60% over the sliding-box method used in the creation of the 1XMM catalog.
@article{arxiv.astro-ph/0603365,
title = {Matched Filters for Source Detection in the Poissonian Noise Regime},
author = {Ian Stewart},
journal= {arXiv preprint arXiv:astro-ph/0603365},
year = {2009}
}