Milkyway Panorama blending.

@mabula-admin I have tried cropping all problematic looking edges in the entire mosaic, and omitting any panels with questionable star shapes, but that still didn't work. I can try as you suggest and omit one panel at a time in that region... Will keep you posted. 

@mabula-admin I've done some testing.

I tried stitching only the frames with Rho Ophiuchi which seems to be the only problem spot in the larger mosaic. 

Using beta 6 I was able to finally get perfect alignment in the stars:

but as you can see, the blending/normalization is broken in beta 6. I know this because in beta 4 the blend was very good (despite bad alignment). 

I even rolled back to the non-beta version (1.083.4) and it normalized very well but the alignment was completely screwed up. This further makes me think this is not an issue of my data as much as the program itself. Let me know what you think.

(I used the exact same settings, the ones you recommended, for both of these except I think I used different projection methods for beta vs non-beta but that shouldn't matter for alignment or blending).

EDIT: Update, I installed beta 6 again to try this all once more... 

now I just can't recreate the perfect alignment (with bad blending). Not sure what I'm doing wrong here. I am using these settings (which you recommended)

I'll download your data and take a look and see what I can figure out.  I have been doing a few mosaics and they can be pretty difficult to get together, especially if you don't have good, consistent data across all panels (all panels shot in similar sky light pollution, sky elevation, time of night etc - I mean if you are shooting across one night just the change in elevation of the object as it moves through the sky will make a difference to the dispersion in the data which will effect the quality across panels, if you shoot over multiple nights that will become even more of a challenge).  Frankly I don't see any better software out there to put it together especially if what you want at the end is a raw FITS data format image which you can then further work on.   I'll let you know how I get on 

@mabula-admin Any chance you had time to look at this? I can't figure out why the Rho region won't stitch properly using all the settings you recommended. I tried going through one by one and figure out which frame might be throwing it off, but no luck.

@chris_g_bne Cool, thanks for checking it out! 🙂

Hi @mikeabr, thank you for testing more yourself. I do know that there was a bug in data normalisation present in the beta's which will be fixed in beta7. This bug could have caused the bad blending for sure.

Now with regard to alignment of the mosaic, I will try to test it this afternoon.

Mabula

Hi @mikeabr,

I am busy with your mosaic now, but now I do see a clear possible problem in the mosaic panels that you provided.

1 of the panels has completely different resolutions then the panels directly from the camera's, this panel indicates that it is already a "mosaic" and thus correct optical distortion correction would not be possible for that panel. The reason is that the Field Of View no longer matches the sensor area which is needed to perform optical distortion correction. It is very technical but that might be the problem. I will try to see if all works without that panels now.

It is the panel Rhode_Island_Rho_Panel-Luminance-REMOVEDLP-crop.fits which is located in your problem area, so this might be the issue.

Furthermore, that panel has been processed with LP correction as well which might be a problem as well for normalization if it was not done accurately. (performing LP correction on individual panels for a mosaic will in most cases make the issue harder and not easier).

Mabula

Hi @mikeabr,

I have succesfully created the mosaics using only the data of one camera at a time.

In star analysis I noticed that with 2000 stars, not everything is reigstered nicely, with 3000 it is in these cases.

I purposely did not include the

Rhode_Island_Rho_Panel-Luminance-REMOVEDLP-crop.fits

panel, because I suspect it is the source of the issues.

I will now try to register all frames except the Rhode_Island panel again to see if I can make it work 😉

Mabula

Hi @mikeabr,

Unfortunately, I can not complete the mosaic with all panels that you provided, somewhere there is a problem in APP or the data somehow.

I am investigating further...

Mabula

@mabula-admin thank you for your efforts, although it's too bad you couldn't get it to work I hope this helps APP become easier to work with once you figure out what's going wrong. Also I feel slightly better that at least it wasn't my fault something was going wrong 😅. I can't figure out what would be causing it in the data, especially when I've been able to get it to stitch with certain updates (in smaller batches) but then not at all in others. Something is going on behind the scenes that maybe you'll be able to uncover. Regardless, this is still the best option out there for generating a mosaic of this size. 

Posted by: @mikeabr

@mabula-admin thank you for your efforts, although it's too bad you couldn't get it to work I hope this helps APP become easier to work with once you figure out what's going wrong. Also I feel slightly better that at least it wasn't my fault something was going wrong 😅. I can't figure out what would be causing it in the data, especially when I've been able to get it to stitch with certain updates (in smaller batches) but then not at all in others. Something is going on behind the scenes that maybe you'll be able to uncover. Regardless, this is still the best option out there for generating a mosaic of this size. 

Hi Michael,

Sorry that it took so long, but yes, I will investigate this thoroughly for further updates down the road !

Mabula

@mabula-admin So does that mean this won't be resolved anytime soon? I really need finish this project... you had success with stitching the mosaic with just one set of data at a time? When I tried that I still had problems with misalignment. At this point, I am willing to give up half my integration data to just get this done if that's what I need to do.

@mabula-admin Ran another integration last night with absolutely all bad edges cropped and omitting most frames from the edges to just have less panels to work with. 

On beta 6 using the exact same settings as before (which you recommended) the whole center of the image is misaligned. 

Then I installed beta 7 and it sat at 95% normalization for 10 hours before I canceled the integration.

@mabula-admin any chance there are any updates on this? I'm still trying my best to assemble this mosaic. 

@mabula-admin I can't even replicate your results with the same data and settings. Can you explain where I'm going wrong?

You said you were able to get a perfect result using the first set of data I provided (a7iv session only) Using the following settings: 

3) 2500 stars

4) Reference Frame (panel 4-1, bottom right corner)

Quadrilaterals, scale start 1, stop 10

Dynamic Distortion Correction on

Same Camera and Optics (when using only a single camera's data)

Registration mode: Mosaic

Model: Calibrated projective

projection: rotate 10 degrees, mercator projection with COP -15 horizontal and vertical 15 

5) Advanced Normalization

6) LNC 1st degree, 3 iterations, MBB enabled at 50%

Using these exact settings and the same data I provided to you, I am not able to get a useable stitch. Misalignment plagues the entire mosaic - although this iteration had Rho Ophiuchi almost perfectly aligned, where other areas are misaligned that weren't before. There is not a consistent result in alignment.

Example:

Can you please advise on how you got the alignment correct?

I have also tried to create smaller bits of the mosaic using fewer panels and can't even get them to register - I get this error:

I'm on beta 9. 

Thanks,

Mike

@mabula-admin the software has two bugs.

1. When changing registration/integration settings after having already run an integration, the changes only apply if you manually re-run star analysis after changing the settings. 

2. Disabling "Calibrated projective" completely and entirely fixed all alignment issues. I have flawless alignment now using data from 2 cameras. Wish it hadn't taken me 5 months to figure this out.

For anyone's reference, here are the settings I used to finally put this mosaic together using data from two different cameras and many shooting sessions:

3) 6,000 stars

4) Quadrilaterals, Scale start 1/stop 10, Use dynamic distortion correction, registration mode mosaic, registration model projective (used a reference frame from the very center of my mosaic)

5) mode: advanced 

6) LNC 1st degree, 3 iterations, Multi-band blending enabled at 50%

Will post my processed result later.