Marc Gentile, Frederic Courbin, Georges Meylan
It is anticipated that the large sky areas covered by planned wide-field weak lensing surveys will reduce statistical errors to such an extent that systematic errors will instead become the dominant source of uncertainty. It is therefore crucial to devise numerical methods to measure galaxy shapes with the least possible systematic errors. We present a simple "forward deconvolution" method, \emph{gfit}, to measure galaxy shapes given telescope and atmospheric smearings, in the presence of noise. The method consists in fitting a single 2D elliptical S\'ersic profile to the data, convolved with the point spread function. We applied \emph{gfit} to the data proposed in the GRavitational lEnsing Accuracy Testing 2010 (GREAT10) Galaxy Challenge. In spite of its simplicity, \emph{gfit} obtained the lowest additive bias {($\sqrt{\mathcal{A}}=0.057\times10^{-4}$)} on the shear power spectrum among twelve different methods and the second lowest multiplicative bias {($\mathcal{M}/2=0.583\times10^{-2}$)}. It remains that \emph{gfit} is a fitting method and is therefore affected by noise bias. However, the simplicity of the underlying galaxy model combined with the use of an efficient customized minimization algorithm allow very competitive performances, at least on the GREAT10 data, for a relatively low computing time.
View original:
http://arxiv.org/abs/1211.4847
No comments:
Post a Comment