Construction of Data Set
In order to construct "3-channel" data, we collected the various
single-channel parts and performed a number of merging
operations to generate a data set that is identical to real
3-channel data which we hope to see in the coming weeks
(spring 1997).
We chose the COMA and SA57 field scans for study due to their
proximity to the NGP and for the large number of JHK scans
taken on the night of 95-04-24 and 95-05-03. This memo describes the results
for four SA57 scans, labels 039, 040, 041 and 042 (corresponding to
the actual K-band scan numbers), and scan 056 - 058, corresponding to the
COMA core.
The protocam scans were ``converted'' to 3-channel-like
scans as follows:
(1) Individual H and K coadds were registered to the corresponding
J coadd. Note: Since the J, H, K scans were aquired with nearly identical
areal coverage, this registration procedure resulted in the loss of only
50-90 pixels along the inscan direction and 10-20 pixels in the cross-scan
direction.
(2) Coordinates in the FITS headers were precessed/converted to
J2000.
(3) A list of sources for each scan was bandmerged and the object
coordinates precessed to J2000.
(4) The additional files required by 3-channel GALWORKS were
generated by bandmerging the individual filter components.
Known Problems
There is distortion between the three bands, so the registration is
not perfect, particularly for the ends of scans. The result is a many
-pixel shift for some objects. GALWORKS corrects for this by registering
small postage-stamp size images during processing.