Comparison of psphot detections on images with different S/N

I tested the object detection efficiency of psphot by running psphot on a number of images supplied by Jim. Not all images show the same field.

1) ps1_i_sub.fits/ps1_i_sub_mask.fits/ps1_i_sub_wts.fits

This image was a composite mosaic Jim put together from four MD04 reference stack frames. This composite mosaic does not include the covariance information from the original stacks, which clearly impacts the object detection. The CMF from this run has 7884 PSF objects (5658 with non NULL PSF_INST_MAG), and 1354 xsrc objects. The image with non-NULL mag PSF objects overplotted shows a number of missed objects along with faint objects and objects that do not appear real.

2) i_sdss_sub.fits/i_sdss_sub.mask.fits/i_sdss_sub.var.fits

This is another composite mosaic of SDSS images. The original images had no covariance information, so it is not lost during this processing. The CMF from this run has 712 PSF objects, all of which have non-NULL PSF_INST_MAG, and 169 xsrc objects. All objects detected in this image appear to be real, and there are only a few cases where a real-looking object is not detected.

3) subaru_ip_064_sci_20.fits/

This is a deep Subaru image with mask and variance created by setting the mask to all zero (accept all pixels) and the variance image equal to VAR = SCI + 21.19**2, where 21.19 is a quick estimate of the image noise. The CMF contains 2015 PSF objects (1989 with not-NULL PSF_INST_MAG) and 25 xsrc objects. The low number is almost certainly due to the poor variance image model, which excludes the majority of the faintest objects.

4) MD04.V2.skycell.078.stk.179047.unconv.fits/MD04.V2.skycell.078.stk.179047.unconv.mask.fits/MD04.V2.skycell.078.stk.179047.unconv.wt.fits

This is one of the raw MD04 reference stack frames used in the mosaic above. This retains the covariance information, and the associated CMF file shows a much better correspondance between image and detected objects. The CMF has 16417 PSF objects (16212 with non-NULL PSF_INST_MAG) and 1276 xsrc objects. Although there are a large number of detections, the majority appear to associate with very faint objects.

5) subaru/MD04 unconv comparison

To check how many of the objects detected in the MD04 reference stack correspond to real faint objects, the detection list from #4 was overlaid on the image from #3. Although there are some detections that do not associate with real objects on this deeper image, it does appear that the majority of the faint detections from #4 are real.