Best Bayer Drizzle for OSC (ASI2600MC), and best sequence for lpc/cbg and channel split

PS: Just remembered this post, where Mabula (@mabula-admin) recommended for my Fuji XH2-S (same Sony sensor as ASI2600MC, same 26 megapixel, also APS-C, 600mm lens) to drizzle with scale 1 and droplets 2.0

==> Can I derive correctly to use the same for ASI2600MC with the 500m FRA500 telescope ?!

I integrated the OSC images (7.7 hours have become 7.6 hours, I left out all frame with quality score < 400):

• 0) Adaptive Airy Disc, 5) Neutralise Background: off, 6) LNC 4th, 3 iterations
• Drizzle/Droplet in 3 variants: 1.0/1.0 - 1.5/1.5 - 1.0/2.0

Here the comparison:

1.0/2.0 is quite a bit noisier, (at least) I cannot see more details

1.0/1.0 and 1.5/1.5 are quite similar - but I see a small bit of more details in 1.0,1.0
(Though it's difficult to compare in Mac's preview, as the 2 files have different pixel size.)

I derive from this to go with 1.0,1.0 ...

Hi Michael @michaelacg,

For workflows like this, I think the proof is in the pudding, so you need to experiment for yourself and see what works best for you, or for your data actually. The ideal workflow can actually depend on the object that you shoot.

But I can of course give you some general guidelines / best practices:

• definitely do lpc/cbg before channel split
• csc calibrate star colors can only be done before channel split
• yes, do LPC and CBG on the Ha and O3 channels as well
• But, if you can and want to combine R,G,B,Ha and O3 to 1 composite, you can also simply do LPC/CBG on that composite
• Personally, I would do workflow 2, register and normalize all subs of all channels when making the R,G,B, Ha & O3 stacks. CSC is wouldl do then on making a RGB composite, perform CSC on that to have accurate star color for your final composite.
• Bonus response: yes, still do reg/norm in APP

Regarding all the drizzle options, there is never a best setting that will work on all data. Drizzle performs well or not well depending on many factors, amount of data, quality of data, dithering, undersampling etc... As a guideline, for Bayer Drizzle, you use 2x the droplet size as for mono data. If you drizzle but do not change the scale, so keep it at 1.0x, then use droplets of 0,5-1 for mono and droplets of 1-2 for RGB. Simply said, for Bayer Drizzle, scale 1.0 and droplets of 2.0 are more or less the same as not drizzling, but still debayering the data. The lower you set the droplet size, the more drizzling becomes apparent, but also the noisier the result will be.

So for drizzle in general, you need to experiment really what is a good value for your data and this is very personal I think. Drizzle is always a balance between getting sharpness or noise... so you will get more sharpness with more noise, or less noise with less sharpness and the droplet size controls this. sharper and noisier is with smaller droplets 😉

Hope this helps?

Mabula

@mabula-admin 

Thank you, Mabula.

By coincidence, I just posted a minute earlier the comparison of the 3 drizzle variants.

Re. you advice to do workflow 2, i.e. register and normalise all 5 channels in one run.

• I wanted to do this ... including 6) Integrate ...
• ... and stumbled over the question, which option I shall choose in 0) RAW/FITS:
  - for OSC I learned to choose Adaptive Airy Disc
  - but for the Ha-OIII filter, there is the option "Ha OIII color", which I interpreted as generating an OSC FITS for my filter frames.
• So my conclusion is:
  a. Integrate OSC and filter frames separately (using Ha-OIII extract XX)
  b. lpc/cbg, maybe csc
  c. separate channels
  d. reg/norm all channels
  e. and finally rotate, crop

Is/where is there a misunderstanding?