Mosaic Star Alignment Problems

[niteman1946]

Hi Group,

I'm following Steve Allan's U-Tube video for building a mosaic from more than two panels.  My current setup is using four panels.  And I will ultimately merge six once I've collected enough photons.
I've used Steve's video before with satisfactory results (at least for me).  But I'm now running into star alignment problems as I conduct the initial steps of the protocol.  That is where we do a star alignment with the "working mode" set to Register/Union Mosaic.  I can initially align some of the panels, but others refuse, and claim insufficient matching star pairs.
The attached Drop Box folder includes the following files:
1.  M33 Framing in SGPro -- screenshot of the six panels, their location and number.
2.  Register/Union Mosaic -- screenshot of StarAlignment Process with typical settings
3.  M33-2 -- Fit file of panel no.2
4.  M33-3 -- Fit file of panel no.3
5.  M33-5 -- Fit file of panel no.5
6.  M33-6 -- Fit file of panel no.6
7.  M33-3to6 -- Fit file of merged panels no.3 into 6
8.  M33-2to5 -- Fit file of merged panels no.2 into 5

Items no.7 and no.8 were successfully aligned using the default settings as shown in item no.2.  Unfortunately, that's as far as I could get.
From my previous efforts with mosaics, I found that the more subs used to integrate each panel did have a beneficial affect on how well the mosaic worked and looked.  So I intend to add photons.  But I would like to get past this initial obstacle.

Any and all advice would be appreciated.

Thanks,

Mark

https://www.dropbox.com/sh/rdmqfglh55u0q16/AAAf2USO79fZKnzvtNXo3mYia?dl=0

[pfile]

i would try MosaicByCoordinates. it requires you to run ImageSolver on all your panels, which can sometimes be fiddly, but if you get past that, the results are pretty much foolproof. you load the output of MosaicByCoordinates into GradientsMergeMosaic and you're done.

rob

[gvanhau]

Hello

Besides the alternative given by Rob, you can also try drawing previews of the overlapping regions of each image pair.

Geert

[pfile]

actually, it seems like you've got some pretty significant optical distortions going on which are probably causing this problem. 2nd order polynomial distortion fitting in ImageSolver resulted in images that don't line up properly (2 and 5). i am experimenting with 3rd order polynomials...

rob

[niteman1946]

Thanks Geert and Rob.

Rob,
To my surprise I was able to wade through the Mosaic By Coordinates approach and got good results.  Well, good is a relative term.  If I pursue M33 any further, I'll probably start over again, and chunk the subs suffering from overcast and moonlit skies (most of them).  But the MBC strategy appears to work.

Your comment about 2nd and 3rd order polynomials doesn't surprise me.  But I'm getting even better optical results from this old scope than expected.  I thought there would be more edge distortion.

Geert,
Can you point me towards an example of how to use the previews for the mosaic?  I could not figure out how to connect the dots on this one.

Thanks again Guys.

Mark

[pfile]

that's good news!

i managed to do it the "old fashioned" way by merging 2 with 3 and 5 with 6 first, then merging both into a master mosaic. for every merge i used distortion correction with the maximum number of iterations, and actually for some of the panes it never converged. then i used "register/union separate" mode on all 4 images against the master mosaic and used GMM on those 4 images. aside from bright-stars-on-edge artifacts, the result is pretty good, though there's still enough misalignment that "average" mode does not look good. had to use "overlay" mode.

in all cases the ability to restrict to previews was pretty important - basically if you have previews defined on your images then StarAlignment (by default) only considers stars inside the previews. in general you can use this to "point out" where the overlaps actually are, which helps StarAlignment not get distracted. when doing the final alignment to the 'master' mosaic i made a preview that's about the size of an individual frame and moved it around the mosaic as i registered each of the 4 sub-images.

rob

[gvanhau]

Hello

Rob was way faster than me, and gave you an example on how to use the previews.

Regards
Geert

[niteman1946]

Hi Geert,

Yep, that worked too!

Thanks,

Mark