Learning to process a larger mosaic.....is difficult

8 replies230 views
Christian Bennich avatar
Hey All

As summer draw closer, I set out to try and build a bigger mosaic.  2x7 panels.
Nothing fancy, just a small constellation. 

I have acquired 1 hour of data for each of the 14 panels and then I need another 6 to plug a few gaps.
Each individual frame is good, and all have been processed in the following way:
  • Calibrating, normalizing and stacking individual color channels
  • For each color (R, G, B)
    • DynamicCrop
    • Imagesolver
    • GraXpert
    • SpectrophotometricColorCalibration


Then I have build up the mosaic in blocks of 4 panels (2 rows).
Which have then been joined together etc. 

Individually, the gradient corrected frames seem great. 

When I start to join them together in larger mosaics, It's evident that some things off. 



This is just a "ScreenTransferFunction" stretch of the image. 

I am wondering if it would make sense to apply something like LinearFit to all my images. Looking at statistics for the individual color channels for the different panels, I can see that there are significant differences in the Median in the images. 

I am trying to figure out what I could do, to make the effects you see in the image, a bit less dominant, before I do more processing. . 

Any input on processing larger mosaics would be greatly appreciated.
The square in the "bottom" that's of bad quality needs replacing. 

Thank you
Helpful Engaging
Noah Tingey avatar
Can you post some of the individual frames? I'm suspicious of how truly gradient-free they are.
Well Written
Christian Bennich avatar
Hi @Noah Tingey 

Sure thing, frames are uploaded here. 

https://drive.google.com/file/d/1107OxVukZ9E_08C_Egc9dhynQsqHvR1r/view?usp=drive_link

3 different stacked frames, from different parts of the mosaic. 
They are maked to make it obvious where in the mosaic they are placed. 

Thank you for responding.
Concise
Noah Tingey avatar
Yeah these have gradients still. I don't recommend using graxpert tbh. I passed it through ABE with a function degree of 1 and it produced a noticeably more uniform background.

Christian Bennich avatar
Thx, I will reprocess 😂
Christian Bennich avatar
Thanks alot for replying. 

GraXpert was the lazy choice 😢
Michael Maas avatar
how did you combine the images?  Did you image solve each frame and then use MosaicByCoordinates to build the individual parts of the mosaic and then use GradientMergeMosaic to stich it all together?

Gradient on the individual frames can be trouble some and it sounds like the advice above is good to get the best individual frames into the mosaic process.
Christian Bennich avatar
My process has been to create an individual color-corrected, gradient-corrected and  platesolved RGB frame for each panel. 
Then using MosaicByCoordinates, with typically 2 frames at a time, TrimMosaicTile on both and the PhotometricMosaic to stitch them together. 

I have reprocessed 5 frames yesterday with PhotometricFluxCalibration and MultiscaleGradientCorrection. 
I have not yet evaluated whether that result is significantly better after joining the 5 frames. 
However - it was pretty clear, that the individual frames was more uniform.

I have not been using GradientMergeMosaic. Does that typically do a better job than PhotometricMosaic?
Helpful Engaging
Michael Maas avatar
I think it does and is what I have used.