I finally went back to that set of 3 photos and got a good result. I'm not happy with the methods required. There is probably a better way. I'm open to advice on what to try next time. There should be a better way, so in any case I'll look through the relevant parts of the source code to see what could be made to work better.
There are three basic issues here, two of which were solved with masks. But editing those masks required fixing a flaw in the mask editing code (see my recent new thread). 1) Exposure. Each photo has the correct average exposure for its part of the scene and I didn't want to change that much. But the top photo had shorter exposure than the middle which is shorter than the bottom. So sections from any one photo have good exposure unchanged. But narrow seams would highlight the exposure difference, while broad seams would be garbage due to issue (2). I still think the better general solution would be to first apply an exposure vertical gradient to each photo (in a separate program if necessary). But I didn't find time to try that. A mask worked a lot better than I expected, to force narrow seams in places that don't highlight the exposure jumps. 2) Alignment. I never got good enough alignment for real blending of any sections. My solution to (3) made the misalignment 1.5 times as bad. The masks intended to fix (1) fixed (2) with no extra effort. 3) Distortion. The resulting image was very distorted. I tried most of the projections and some were worse and none were significantly better than the default. After much experimentation, I started without the lens a,b,c,d,e,g,t parameters, then rotated the anchor image -90 and reoptimized (which didn't hurt alignment much), then cut the Hfov in half, and reoptimized which destroyed alignment (I assume because the original Hfov was correct) then added a,b,c,d,e,g,t parameters to get back to decent alignment (I'd really like to know what those are doing and how they compensate for an intentionally wrong Hfov. But I'm already investigating in too many different subtopics at once). Finally, I stitched with Cylindrical and used an external program to rotate +90. Given the distortion this had when simply done sideways with Cylindrical, I expected that cutting the Hfov would do what it did. But I have no clue why it had that distortion. It seems like there should be a less distorting projection without needing to mess with the alignment. -- A list of frequently asked questions is available at: http://wiki.panotools.org/Hugin_FAQ --- You received this message because you are subscribed to the Google Groups "hugin and other free panoramic software" group. To unsubscribe from this group and stop receiving emails from it, send an email to hugin-ptx+unsubscr...@googlegroups.com. To view this discussion on the web visit https://groups.google.com/d/msgid/hugin-ptx/42d0ad7a-1726-441a-9362-ac71505e6728n%40googlegroups.com.
