DSO Mosaics

[MattJenko]

I would like to canvas some opinions on how people manage DSO mosaics in terms of matching backgrounds and also managing gradients and the overlapping areas.

I have recently given a 4 panel mosaic a go, and found the stitching together to be the most challenging aspect of it all. I have a small chipped camera so will be attempting a few more in future as well so keen to learn a few more techniques.

Thanks in advance,

Matt

[ollypenrice]

I tend to do it like this:

Stack the L, RGB and any NB separately for each panel.

Crop off the edges. (Vital!)

Take the RGB into Pixinsight for DBE. From my site I rarely need to run the L or NB through it but I do if necessary. I'll also apply SCNR green.

Give the still-linear panels of RGB to Registar and ask it co-register the set to one from near the middle. 

Run Registar's 'calibrate images' routine.

Save the registered/calibrated sets in case a fix is needed later.

Ask Registar to combine the set.

Take this into Ps for a test stretch to see if there are any lines showing. If there are they need to be covered over using one of the saved individual panels which does cover the line. This often isn't necessary but is easy in Layers.

Return to Regsitar and co-register the L and NB layers to the new full RGB and make full mosaics of L and NB which will now fit perfectly onto the RGB.

In the case of very large areas of sky with attendant problems of field curvature you need to make a template and register individual panels to this. I believe you can use PI to make a template from its star charts but I haven't done this. I've used an upsized camera lens widefield (upsized to the final size of the proposed mosaic and then discarded.)

Once the linear RGB mosaic has been constructed you could run it through DBE again (and maybe Background Neutralization) to help with the global balancing. On our big Orion there was a lot of work needed to pull the global CB into shape.

Olly

[swag72]

I use Pixinsight for my mosaics - This is the process I follow.

Works perfectly and gives me no seams and gradients

[MattJenko]

Thanks Olly. From that description it appears that you lean heavily on Registar to do the bulk of the grunt work in terms of the stitching once the individual images are initially stacked and colour calibrated. This was exactly the stage I had trouble with using PixInsight, although I got something which worked ok in the end, but it was the double cluster so I could be quite aggressive with backgrounds. I am sure I would not get away with what I did for a more continuous nebula based mosaic, so I'll have a look at Registar.

[MattJenko]

Sara - that is perfect, thank you.

[ollypenrice]

Another Registar bonus is that you can ask it to resize and re-curve a long FL image to match a widefield. Combining them is best done in Ps because it depends on a degree of artistry since a blunt dropping-in of higher res data looks very obvious.

I must watch Steve's video again.

I do think that running DBE on the original linear panels is effective, though, for getting backgrounds as close as possible before stitching.

Olly

[MattJenko]

I had a go using Sara's linked method and it was much easier than what I did first up, which was to use Crop to create a large panel to register the images to, which meant I couldn't accurately use the linear fitting tools that this video's method allows. I had to use a fairly large feather region when using GradientMergeMosaic, but the end result was better than my longwinded effort, and much less manual and time consuming.

I will still look at Registar though in due course - thanks again, learnt a lot in the last few days. The learning never stops