I’m curious about how the new Stitch software calculates its Vignette Correction. I’ve used Photoshop’s vignette removal in my Gigapan workflow for quite a while, and I know that it uses camera/lens profiles that it matches to the images based on their metadata or a manual selection. Is Stitch inferring the amount needed from analysis of the image, or is there some other secret sauce?

Yes, the software first does alignment to determine how the images overlap, then it uses the corresponding pixels in the overlap regions between horizontally- and vertically-adjacent images to infer a vignette attentuation function.

This way of doing things is nice for several reasons:
(1) It works when the input images don’t have EXIF for the lens (you’d want to know lens model, focal length, and aperture, at minimum),
(2) It does not require a large library of vignette attenuation data indexed by lens model, focal length, and aperture, which would be incomplete every time a new lens comes on the market,
(3) It does not require special rules when a tele-extender or filter causes a change in the vignetting,
(4) It’s automatic and transparent to the user!

This analysis only adds a few seconds of time to the entire stitching process.

If you look in the Stitching Notes, you’ll see two coefficients listed that are part of the attenuation function that is inferred. It is circularly symmetric and it’s a quartic polynomial.

-Paul
GigaPan Stitch Developer

Thanks Paul – I’ll have to give it a try! It will certainly streamline my workflow.