Status: Won't Fix
Fix Version/s: None
(See also HSC-1025)
One poor-man's approach to deblending multiple bands and visits is
- Deblend in one band (or a combination, e.g. chi^2 band) using an SDSS-like algorithm that produces templates for each child
- Take the templates (or possible model fits to those templates) and use them to deblend the objects in all other bands
It will probably be necessary to include undetected objects in this fit (cf. HSC-1023), using point-source models.
Note that this does not allow for the different seeing in each band. This is not obviously catastrophic for reasonably wide blends as the total flux in each pixel is conserved, and if only one template is important near an object's centre the exact form of the template is unimportant.
In discussion with Jim Bosch the consensus is that it's too early to consider this issue, and that 40 SPs makes it too unwieldy to schedule. Closing as Won't Fix for now; we should break it up and address it as small stories in future.