Gradient Merge Mosiac Video Tutorial

[ugatza]

I answered myself. I've discovered I need more area of ??overlap between different mosaic tiles.

[Juan Conejero]

Hi ugatza,

I answered myself. I've discovered I need more area of ??overlap between different mosaic tiles.

I'm sorry, I overlooked your question. Indeed, this message means that SA cannot build an initial list of possible star matches between both frames. This usually happens because the overlapping area is too small, but it can also be due to other reaons: too large scale differences, one of the frames has bad focus, too strong local distortions, etc.

When this problem arises, you can solve it by defining previews on both images covering (roughly, you don't have to be very accurate here) the overlapped area. When SA detects the previews, it restricts the star matching task to those areas, which removes the uncertainty that is causing it to fail. Please let me know if this works for you.

[ugatza]

Hi, Juan

I used the trick you told me and it works  , although it is true then I tried later without it and has also given me good results. Previously I could not make or the good fit of all the tesserae, or that after alignments with the final mosaic pattern, and then join me the Gradient tool Mosaic Merge as videotutorial of Catanonia explains. However, now that you have done everything correctly.

By the way, what is better? • Process each of the tiles in terms of what you have inside (in a different way if there are only stars or nebula only) or first make the mosaic and then process the area as a whole?

[georg.viehoever]

By the way, what is better? • Process each of the tiles in terms of what you have inside (in a different way if there are only stars or nebula only) or first make the mosaic and then process the area as a whole?

The short answer: I don't know:
The long answer:

• You certainly should do the calibration and debayering on separate frames (so do not merge bias, dark, raw lights or un-debayered frames)
• The arithmetic of GradientMergeMosaic is much more accurate than anything that common panorama tools for daylight photography do. It also preserves most of the linearity that pictures have. This would mean that you could merge frames immediately after calibration. This is what I usually did for my rather modest mosaics.
• On the other hand, the computations are quite complex, and may introduce artifacts of their own. So you may want to wait with merging until after you have removed noise and improved the actual signal.
• Another consideration is that processing the merged mosaic is much more memory/CPU intense than doing a single frame. So you may want to process a single frame first, then apply the same steps (maybe with minor tweaks) to the other frames, and do only the final touch-up on the merged result.

Maybe others who have used GradientMergeMosaic for their images can share their views on this.

Georg

[Oleg Astro]

...
When this problem arises, you can solve it by defining previews on both images covering (roughly, you don't have to be very accurate here) the overlapped area. When SA detects the previews, it restricts the star matching task to those areas, which removes the uncertainty that is causing it to fail. Please let me know if this works for you.

Hi Juan,

I've followed your recommendation for the resolution of the similar problem.
It worked!

[alvinjamur]

This is an incredible video!!!!!
PI should definitely make this available as an "official video" for
mosaic workflow after solving the sound issue.
Absolutely fantastic!!

- aLV

[chris_todd]

Thanks for the awesome tutorial!  I'm both a PI and astrophotography newbie, but I understood everything you did and why you did it.  Well done!

One question, though: if you're doing a large number of panels, and don't have a machine as beefy as yours, will it simply take longer as everything gets swapped to disk, or is there some minimum memory and processor combination below which you just shouldn't bother?

I ask because I have a 2.5GHz MacBook Pro with 4GB RAM, and am contemplating trying a 12 or more panel mosaic at the Texas Star Party next month.  I wouldn't mind letting it run overnight, but I want to be sure it will actually finish.   

[georg.viehoever]

4 GB is certainly sufficient for a 3 panel mosaic with 10 MPixels RGB each. But beyond that I dont know, 12 panels are unlikely. Try yourself by adding more and more panels, and then extrapolate the runtime. If in doubt, rent a larger machine, e.g. via Amazon EC2 ($2 or so/hour for 64 GB machine).

Georg

[Graham]

Hi!

Interested in the mosaic functions however I am unable to get the video to play. Is it available as a straight download? How long do you have to wait to watch it and can I watch it off line?

Thanks,

Graham.

[georg.viehoever]

You should be able to download it using the context menue (right mouse button), "Save target as...". If your player has difficulties with .mp4, the VLC player http://www.videolan.org/vlc/ . There is also a video on youtube http://www.youtube.com/watch?v=H2r3ZYqCaP8 that should give you some information.
Georg

[Graham]

Thanks Georg, I should have thought of that.

Cheers.

[cgeib]

Broken Link. Video no play.

[shurik]

Does anyone have a working link to the videos in this thread, none of them are working right now

[msmythers]

This is the video your looking for. I have a copy of the original and they are the same.

https://www.youtube.com/watch?v=H2r3ZYqCaP8


Mike