I just stumbled over a nasty Problem with dust on my lens / sensor. I took a Gigapixel with about 2.500 Images. On all of them are "dust on sensor" pollution. Now Autopano had a pretty hard time stitching the Gigapixel because of the always repeating pattern of the dust. Would it be possible that APG detects such dust spots (as they are on the same position on every Image) and ignores them for stitching?
Don't know how hard this would be to implement - but I guess I'm not solely with this problem! ;-)
I had a familiar issue with dust.
A) remove it before stitching. If the issues rise mainly in the sky you can automate it in Lightroom or PhaseOne (donÂ´t automate it with ground-images: itÂ´s an automated clone-thing and so it doubles a spot)
B) try PTGui - this one looks for object-structures in ajacant images. APGÂ´s finding of CPs sometimes seems to be rather funny. Usually dust is at least a bit unsharp and is of less structure than some objectÂ´s-structures.
It helped me a lot in this pano: http://360impressions.de/MKP_Panorama/
. . . After importing via XML (which was near-perfect) by optimizing APG set CPs on the dust-spots on the ceiling between the clear structures while PGTui used the clear structures for optimizing CPs.
Result: PTGui did the job.
In my eyes - to make it clear - APG is the better stitcher. But sometimes it fails completely - the reason in my eyes is the optimization: it must be adjustable. We must be able to set something like a "catch radius" to avoid that it looks in a too wide range and puzzles nearly perfect set connections by xml with placing "optimized" CPs in very funny regions.