0

I have a large series (~1000) of wide-angle photos, taken with a medium quality digital camera over a period of 2 hours. I'd like to automatically figure out the rotation speed and center of rotation (north star), and using that info and the photo's EXIF timestamps, stack all the photos into a single, simulated long exposure photo like the pros do with their fancy motorized telescopes.

I've tried StarStaX, but while that does nice median calculations and creates pretty star-trail photos, it doesn't do alignment. Hugin's cpfind is better at real-world control point detection, and dies when trying to pinpoint star location.

I think I can do a pretty good job of identifying stars using local brightest points, but I don't remember enough freshman math to take the ~500 vectors aligning the stars between two images and boil that down to a center of rotation and a theta.

Any suggestions? Image stabilization code?

Benjamin H
  • 5,164
  • 6
  • 34
  • 42

1 Answers1

0

Can you post link to 2-3 samples of your star photos? I guess optical flow calculation and then affine transformation (rotation and possible shift) search for this flow will help.

Vit
  • 793
  • 4
  • 17
  • Absolutely. [Zip of a few frames](https://drive.google.com/file/d/0ByXCDxKWoabVRVN6OS1HZHBjLTQ/edit?usp=sharing) – Benjamin H Apr 29 '14 at 02:41
  • Wow!GoPro Hero3 night sky jpeg photo.It is hardly "medium quality", i'd say "low quality as it is" due to Sony IMX117CQT 1/2.3 tiny sensor and nightmare 1.55 μm unit pixel. I think it is not worth to waste your and others time trying to fix unfixable things. I have tried change levels and could see only 4 bright dots among high level of noise. It is hardly enough to calc correct transformation for alignment. Try reshoot it using better sensor (at least APS-C-sized) and don't compress it to jpeg (lossless png is better). Then I will try to help. GoPro is for sunny day video, not for night foto. – Vit Apr 29 '14 at 19:52