A robust algorithm for the pointing refinement and registration of astronomical images

We present a generic algorithm for performing astronomical image registration and pointing refinement. The method is based on the matching of positions and fluxes of available point sources in image overlap regions. This information is used to compute a set of image offset corrections by globally minimizing a weighted sum of all matched point-source positional differences in a prespecified reference image frame. A fast linear sparse matrix solver is used for the minimization. From these corrections, the pointings and orientations of images can be refined in either a relative sense in which pointings become fixed ( registered) relative to a single input image, or in an absolute sense ( in the International Celestial Reference System [ICRS]) if absolute point-source information is known. The latter provides absolute pointing refinement to an accuracy that depends on the robustness of point-source extractions, match statistics, and accuracy of the astrometric catalog used. The software is currently used in the Spitzer image-processing pipelines, although it is adaptable to any astronomical imaging system that uses the FITS image format and world coordinate system (WCS) pointing standard. We test the algorithm using Monte Carlo simulations and compare them to image data acquired with the Infrared Array Camera (IRAC) onboard the Spitzer Space Telescope. We find that dispersions in matched source separations after refinement are entirely consistent with centroiding errors in source extractions, implying that systematic uncertainties due to inaccurately calibrated distortions are negligible. For these data, we predict refinements to better than similar to 70 and similar to 280 mas ( 2 sigma radial) for the IRAC 3.6 and 8 mum bands, respectively. These bands bracket two extremes in available source matches, and for the data under study, correspond to an average of about 55 and 8 matches per image in these two bands, respectively.