First of all, I think that Xpano is great: it’s simple, and efficient.
However, on close inspection of the results, I observe a lack of detail compared to the original images, and also compared to what hugin creates. Almost as if there was some slight haze added.
Below are 2 screenshots comparing Hugin (top) and Xpano (bottom) stitched outputs:
Pay attention to the trees, this is wgere the “haze” seems most evident:
Interesting, after some pixel peeping, I can also see it on some of my examples, I’ll try to investigate.
Thanks for sharing the comparison!
Not sure if it is the same effect, but I also observed that the panorama created by xpano is somehow more dull than the one from hugin.
In this case, the exposure of the input images varies a lot, though (I had forgotten to use manual mode…), maybe this causing the problem? I think I read somewhere that Hugin is supposed to be more capable in adjusting the exposure.
Here are the original images in case they are useful for debugging (license: CC BY-NC-SA)
Thanks a lot for the images, great for testing!
The algorithm can indeed get thrown off with inconsistent exposures. In this case I can get a better result from Xpano by excluding the DSC00862.JPG image from the stitch (by CTRL clicking the thumbnail), but there is still an exposure artifact remaining in the middle of the right edge,
A user suggested to use integrate Multiblend (Seamless blending · Issue #76 · krupkat/xpano · GitHub) which could improve the situation, until then I recommend locking the exposure for your photos
Just for information, in the case of the panorama shown in my OP, all the frames had been shot in manual mode, with the same exposure settings.
I have done some tests and the results with Xpano + Multiblend are quite nice so far, although not perfect:
I’ll look at integrating this process to the next version.