24mm Widefield Imag...
 
Share:
Notifications
Clear all

15th Feb 2024: Astro Pixel Processor 2.0.0-beta29 released - macOS native File Chooser, macOS CMD-Q fixed, read-only Fits on network fixed and other bug fixes

7th December 2023:  added payment option Alipay to purchase Astro Pixel Processor from China, Hong Kong, Macau, Taiwan, Korea, Japan and other countries where Alipay is used.

 

24mm Widefield Images Not Registering for Mosaic

5 Posts
2 Users
0 Likes
3,329 Views
(@aeroman)
Molecular Cloud
Joined: 6 years ago
Posts: 3
Topic starter  

I am using the trial version of the software to see if it can help me with super large mosaics of the Milky Way before committing to purchasing.  I have 17 panels I will ultimately be stitching together from my Canon 5DM2 and a Rokinon 24mm lens.  Almost all of the panels have been processed with PI and are ready to be stitched together.  I have PTUgi that does a decent job stitching the images together, but I am trying to get the stitch with regard to color a little better matched between the panels and someone suggested APP.  I have good overlap on images (at least 15%) and tens of thousands of stars per panel.  In using APP, I've managed to get the program to identify at least 15,000 stars in both panels to be merged, but every time I try to register for the mosaic, it tells me there are not enough stars to match the reference image.  Other programs can stitch without a problem, so I know there is enough overlap.  I have tried finding a solution on the net and have tried about a dozen different setting variations with no luck.  Any suggestions?


   
ReplyQuote
(@mabula-admin)
Universe Admin
Joined: 7 years ago
Posts: 4366
 

Hi @aeroman,

That sounds like there should be plenty of stars to get this working properly in APP.

First of all, detecting 15000 stars is very likely overkill and will only increase registration time. I would advise to lower it to a value of about 3000-6000 first.

Secondly, can you share which settings you actually used in 4) Register ? Depending on the settings, different mosaic algortihms are used so it's important to know the exact settings here.

Did you select a reference yourself or did you let APP choose? Have you tried using another reference frame?

Have you enabled dynamic distortion correction and have you increased the scale stop?

Your 24mmRokinon objective might have strong optical distortion, did you try increasing the distortion margin to 0.1-0.2 ?

Finally, your field of view is probably very large, did you use the calibrated projective model?

As a side note, preparing your panels in another application should be no problem unless you have used thin plate splines in the registration model in the other application, that could actually be destructive to the geometry in your data. Thin plate splines will project & bend your data arbitrarily and as such is not a very good solution. If you used a projective model, then this should be no problem at all.

Kind regards,

Mabula


   
ReplyQuote
(@aeroman)
Molecular Cloud
Joined: 6 years ago
Posts: 3
Topic starter  

Thanks for the reply Mabula.  I was watching your video about making mosaics and I missed the step for choosing the reference image (the sound doesn't work on the video).  Regardless, I have reduced the star count as you suggested and I chose a reference frame, but it seems to either be taking a long time or is hung up on "setting new reference frame..." and I can't perform any other actions.  CPU usage is very low, so I can't tell if the program is actually processing anything at the time and the console is not updating and progress.  The frame is centered around Deneb, so there is a lot of the Milky Way in the image, so I don't know if that is a problem or not.  Thanks for any assistance you can provide.  Once I get through the reference frame process I'll be able to try some other suggestions.


   
ReplyQuote
(@aeroman)
Molecular Cloud
Joined: 6 years ago
Posts: 3
Topic starter  

Okay, the program just got stuck and I restarted and it appears to have selected a frame as the 'integrated ref'.  I am still having problems with the registration.  In all tests I have left the pattern recognition to 'quadrilaterals', but have varied the scale stop between 8 and 15.  I have a fair amount of overlap between the panels so that shouldn't be the problem.

As there is some rotation between the panels since I imaged on an equatorial mount, I've tried flipping the descriptors or not for each variation.

I done most of the testing with dynamic distortion correction, with just a few tests with same camera and optics.

I've tried the distortion margin at 0.1 to 0.4.  In truth, the images were already corrected pretty well for distortion by other processing beforehand.

Registration mode is set to mosaic, and I've used calibrated projective for all tests.

Still no luck registering.


   
ReplyQuote
(@mabula-admin)
Universe Admin
Joined: 7 years ago
Posts: 4366
 
Posted by: aeroman

Okay, the program just got stuck and I restarted and it appears to have selected a frame as the 'integrated ref'.  I am still having problems with the registration.  In all tests I have left the pattern recognition to 'quadrilaterals', but have varied the scale stop between 8 and 15.  I have a fair amount of overlap between the panels so that shouldn't be the problem.

As there is some rotation between the panels since I imaged on an equatorial mount, I've tried flipping the descriptors or not for each variation.

I done most of the testing with dynamic distortion correction, with just a few tests with same camera and optics.

I've tried the distortion margin at 0.1 to 0.4.  In truth, the images were already corrected pretty well for distortion by other processing beforehand.

Registration mode is set to mosaic, and I've used calibrated projective for all tests.

Still no luck registering.

@aeroman,

First of all, apologies for the long response time.

You indicate that your frames were already pretty well corrected. How did you do this? Like I indicated, if you used an application that uses thin plate splines, it could very well be the reason for failure here. If you didn't do this, I wonder how you did correct the frames for optical distortion.

It would not be the first time that panels registered with another application fail in APP because the other aplication was destructive to the geometry of the data.

Do you still have the original calibrated light frames of the indivdual panels?

Probably, if you want, I can have a look at your panels and verify why it is failing. If you have 15% overlap, it should work, so I would be happy to give it a try.

Kind regards,

Mabula

 


   
ReplyQuote
Share: