A METHOD BASED ON WAVELET TRANSFORMS FOR SOURCE DETECTION IN PHOTON-COUNTING DETECTOR IMAGES .2. APPLICATION TO ROSAT PSPC IMAGES

We apply to the specific case of images taken with the ROSAT PSPC dete ctor our wavelet-based X-ray source detection algorithm presented in a companion paper. Such images are characterized by the presence of det ector ''ribs,'' strongly varying point-spread function, and vignetting , so that their analysis provides a challenge for any detection algori thm. First, we apply the algorithm to simulated images of a flat backg round, as seen with the PSPC, in order to calibrate the number of spur ious detections as a function of significance threshold and to ascerta in that the spatial distribution of spurious detections is uniform, i. e., unaffected by the ribs; this goal was achieved using the exposure mag in the detection procedure. Then, we analyze simulations of PSPC i mages with a realistic number of point sources; the results are used t o determine the efficiency of source detection and the accuracy of out put quantities such as source count rate, size, and position, upon a c omparison with input source data. It turns out that sources with 10 ph otons or less may be confidently detected near the image center in med ium-length (similar to 10(4) s), background-limited PSPC exposures. Th e positions of sources detected near the image center (off-axis angles <15') are accurate to within a few arcseconds. Output count rates and sizes are in agreement with the input quantities, within a factor of 2 in 90% of the cases. The errors on position, count rate, and size in crease with off-axis angle and for detections of lower significance. W e have also checked that the upper limits computed with our method are consistent with the count rates of undetected input sources. Finally, we have tested the algorithm by applying it on various actual PSPC im ages, among the most challenging for automated detection procedures (c rowded fields, extended sources, and nonuniform diffuse emission). The performance of our method in these images is satisfactory and outperf orms those of other current X-ray detection techniques, such as those employed to produce the MPE and WGA catalogs of PSPC sources, in terms of both detection reliability and efficiency. We have also investigat ed the theoretical limit for point-source detection, with the result t hat even sources with only 2-3 photons may be reliably detected using an efficient method in images with sufficiently high resolution and lo w background.