EMphot: photometric software with Bayesian priors. Application to GALEX.
Photometry of astrophysical sources, galaxies and stars, in crowded field images, if an old problem, is still a challenging goal, with current and future survey missions releasing new data with increased sensibility, resolution and field of view. Photometry of extended sources has been addressed several times using various techniques: background determination via sigma clipping, adaptive-aperture, point-spread-function photometry, isophotal photometry, to lists some.
Our purpose is to estimate the flux in a low resolution band using prior information (position and shape) from a better resolved band, in a Bayesian approach under the Poisson noise assumption. The solution is reached with an Expectation-Maximization (EM) algorithm for solving the photometry and includes several steps: prior shapes deblending in high resolution images, astrometry correction, PSF optimization, background correction from the residual.
We apply this software to the Deep Imaging Survey (DIS) of the GALEX mission, which observes in two UV bands with long exposure times (~ 70'000s), and produces deep sky images of 1 square degree, with hundreds of thousands of galaxies or stars. Priors are computed from CFHTLS data. These UV observations are of lower resolution than same field observed in visible bands, and with a very faint signal dominated by the photon shot noise, with background level around 100 (resp. 10) counts in the near (resp. far) UV band.
The resulting photometric accuracy is quantified with both completely simulated crowded fields and simulated objects added on top of the real images. Finally, compared to blind photometry estimation, the method leads to small and flat residual, increases the faint source detection threshold and provides a better accuracy for bright contaminated objects. On the processed DIS fields, EM provides good photometry and completeness down to magnitude 25.5, which is 1 magnitude deeper than the GALEX pipeline.
Return to poster list