The "mxdn" score refers to the peak flux of the star normalized by its integrated flux. Stars have "hard" profiles (very peaky compared to their total flux) and extended sources appear "soft".
The scoring threshold is 2.0, but only one band has to achieve or exceed this value to be included in the extended source database (this is why many sources can have "mxdn" scores less than 2.0; see plot).
There are many faint "unknown" sources, but these have integrated fluxes typically beyond the level-1 specifications (K = 13.5 or so). They have no classification because they are too faint on the POSS or DSS to discern their nature.
The bright false detections (red triangles), K < 13.0, arise from the transient seeing change in scan 052; see Coma Seeing and the relevant seeing plot Seeing, scan 052.
The "mxdn" score distribution of the galaxies in J and H look very similar, but at K the scores are typically much lower -- this is not well understood at this time.
"sh" Score
The "sh" score refers to the extent of the inner radial profile. Galaxies are well separated from stars in thei parameter space, but not well separated from duble stars and other "complex" non-extended sources.
Similar to the "mxdn" result, the K distribution looks peculiarly different from the J and H distributions.
"wsh" Score
The "wedge" shape is designed to minimize contamination from double stars. The plot shows that most of the false detections have rather small scores (<10), particularly the few double stars. But also note that many galaxies have small wsh scores in the K band, but forturnately not in the J or H bands (thus their inclusion in the database). Again, the K band scores are much smaller than the corresponding values in J and H.
"r23" Score
The "r23" score is very effective against triple stars. Since Coma is located near the pole, very few stars (and even fewer multiple stars) are in play. It is most important that this score be used cautiously under these conditions (else we lose real galaxies).
"trip" Score
The "trip" score is a weighted combination of the "wsh", "r23", "vmean" and "vint" scores, with "wsh" having the largest weight. Although the four scores are slightly correlated, the combination does provide a convenient measure of the "extendedness" of an object. However, as with the r23 score, Coma does not really require this score due to the low stellar number density.
Conclusion
There are very few false detections in this data set, with the bulk of the bright false detections due to untracked seeing in a rather unusual "event" occuring in scan 052 on the night of the 23rd. All of the scores shown here, including "mxdn", "sh", "wsh", "r23", and "trip" show a puzzling trend: the K score distribution is very different from that of J and H. The K scores are typically much lower than the J and H scores. Is this simply due to the increased sensitivity in J and H relative to K? Is it related to the poor sub-pixel dithering associated with the K-band data? Related, are the scores depressed by the increased scatter in the K shape ridge? These are open questions at this time.
