Problems creating mosaic

andefeldt

Member
Jan 25, 2020
19
1
Hi,
I'm trying to create my first mosaic of M31 and it's surrounding space. However, I can't get a nice looking image. It's not seamless and it has dark or multicolored spots in the area around some of the stars in the overlap between two images.

I've tried running combining the two bottom images, then the two top images and then combining top and bottom. I've also tried combining top and bottom on the left side, combining top and bottom on the right side and then combining left and right.

In GMM I've tried multiple shrink and feather settings. I can get decent results when combining eg. the entire left panel and the entire right panel. It gets bad when trying to combine the left and right panel.

My workflow is:

Calibration, CC, debayer and register each panel.
Crop and DBE each panel
SA (Register/Union Mosaic with Frame Adaptation) on 2x2 and ending with creating a rough 4 panel mosaic (multiple gaps)
SA (Register/Union Separate with Frame Adaptation) on 2x2
GMM (various tries with shrink and feather settings) on the recent registered panels to create the image.

What am I doing wrong? I've spend wayyyy too many hours on this today and I'm just about to give up on the project.

I've uploaded my individual cropped and DBE'ed files. Could some of you take a look and try and create a mosaic and show me how??

Files are here:
 

pfile

PTeam Member
Nov 23, 2009
5,689
116
first thing i can see is that there's enough optical distortion that just plain registration won't work. one way or another you're going to have to do some distortion correction. what i did was to use ImageSolver with distortion correction turned on, then use MosaicByCoordinates in high-quality mode. that

secondly something is weird with the left hand panels. the core of the galaxy has a blue halo and that's not right. subsequently if i DBE these images and use MosaicByCoordinates + GMM, the mosaic is generated but it has a strong brown cast on the left and a blue cast on the right. somehow these images are not compatible with one another, and i'm not sure why. but the galaxy core color must be a clue.

messing with the feather controls i was able to get most of the stars to stop "blooming" but there are a couple more left. usually you can take care of those by deleting the bright star from one or more of the panes that overlap and keeping it in just one pane. you can use clonestamp to do this.

finally i ran into a problem - it seems that DNALinear fit no longer works with PI 1.8.8-6. this is a pretty important tool for mosaics since it lets you linear fit images based on their overlap without any messing around. linear fit itself can only work on images which are registered to one another (meaning, identical images of the same field, not a mosaic), so what DNALF does is compute the intersection of two images, and figures out the linear fitting between them.

rob

Screen Shot 2020-09-23 at 10.17.45 PM.jpg
 

andefeldt

Member
Jan 25, 2020
19
1
Hi Rob,

When I used SA I enabled Distortion Correction (can't remember the exact name as I'm no where near a PixInsight installation).

I wondered about that bluish color of the core in the left panel as well. 3 panels are from the same night, the last from the week after. They are all 30 x 180s captured with a IR/UV cut filter in front of the ZWO 294MC Pro camera cooled to -10°C, gain 120 and offset 19. Could it be because of some of the new features in the calibration, alignment and stacking processes? Maybe I did something wrong here?

But wow, great effort. I'll try copying your progress. But what to do with the color and the DNAlinear problems?
 

rakla

Well-known member
Dec 17, 2018
45
7
Switzerland
ko-fi.com
Maybe give this approach a try -> https://pixinsight.com/forum/index.php?threads/mosaic-with-3-panels-failed.15107/post-91837

It might not solve all your problems but gives you another approach compared to GMM. It uses some of the modules suggested by Rob.

dnaLinearFit (v0.4) appears to be working for my data with the latest release of PI, but I had situations where it returned an error (NaN) if the data of a panel in question was very different (weak signal?) compared to the reference panel.

/Ralf
 

pfile

PTeam Member
Nov 23, 2009
5,689
116
@rakla i might be running an older version of DNALF - need to check. good to hear that 0.4 is working.

@andefeldt DBE can sort of help with the color of the background but because that cast continues thru the galaxy i ended up with the colors still present near the edges of the galaxy. i am really not sure what's going on; one thing to try might be to split the channels first before making the mosaic, just to try and get a handle on whether or not there is a problematic channel, but of course that's a lot of work.
 

andefeldt

Member
Jan 25, 2020
19
1
Hi,
I recalibrated, corrected, aligned and registered the two panels on the left side (with the wrong colored core). This time they came out without weird colors. No idea why the colors were wrong last time. Now it looks much better and I can use the StarAlignment with the mosaic options and then GMM without problems. Now I just need to do a correct DBE etc. and then tweak the settings in GMM. A very quick and dirty version came out soo much better than before :)

I just tried the ImageSolve + MosaicByCoordinates and then GMM on a 2x3 mosaic I shot a month ago but gave up upon as the overlap was too small (5%), or so I thought. This combination of Scripts gave me a very nice and very big mosaic - again just need to tweak the various settings a bit.

I'm one happy camper now!

Thanks for your help and great suggestions.
 
  • Like
Reactions: pfile

1DegreeN

Active member
Jan 12, 2014
26
1
Check out the recently released PhotometricMosaic script. You'll find references to it in the forum, it has to be downloaded and added to PI's scripts. I've tested it just once but it is really easy to use and created an excellent result.