Hi, Markku Kolkka wrote: > Don't use stacks with hand-held images. Hugin believes that > images in a stack are pre-aligned perfectly.
Explicit stacks do help with some other things though (cpfind+align_images, whole-stack masking etc.) Just remember to uncheck the "link" checkbox in the images tab for all the stacks (it's in the image orientation section). This allows images to be aligned within a stack, at least in angular space. (There seems to be a bug with XYZ alignment not being properly unlinked even though the optimizer assumes it is, but I haven't investigated that yet to report it, and XYZ isn't an issue for wide panoramas anyway. XYZ does help a little for parallax on flat-ish ground with less than 180deg though, so in that case I unstack the images after creating control points.) My method (working hand-held) is to take the photos with the camera settings basically as you do. Then in Hugin 2010.04: 1) Set up *unlinked* stacks. 2) Mask moving people etc. 3) Temporarily mask out any obviously parallaxed foreground. (Assuming that you prefer the background to be properly aligned.) 4) Run "cpfind (multirow/stacked)" on all. As far as I can tell this uses align_image_stack within stacks and normal cpfind between stacks. Cpfind seems to be aware of the masks, but align_image_stack isn't - deal with that in step 5. I then remove bad control points: 5) Edit->Remove control points in masks - this eliminates the masked out foreground and moving objects. Remove the temporary foreground masks now. 6) Run Celeste if there are clouds. 7) "Clean control points" button to remove outliers. 8) Edit->Fine tune, then remove points with <0.8 correlation. <0.9 works well if you only use align_image_stack, but cpfind creates lots of points with lower correlations for me. 9) During optimizing, I remove control points in stages between rerunning the optimizer, based on their mismatch distance (e.g. >10, >5, >2, >1). Sometimes I need to manually add points between stacks. After the initial positions-only optimization and straightening and centring the preview, I use "calculate optimal size" in the Stitcher tab so that the pixel mismatches are accurate for successive optimization passes. For stages 8 and 9 use "View->Control point table" to select and remove points with low correlation and high pixel mismatch. Then I mask lens flares etc., do photometric optimization and blending as described. If the foreground looks too bad I crop it out. It's a bit overly complex, but for hand-held panoramas taken in the wind at the top of a mountain, computers just aren't yet clever enough to automate the process fully... I too welcome any suggestions of course, but this method at least is working well for me. Gareth -- You received this message because you are subscribed to the Google Groups "Hugin and other free panoramic software" group. A list of frequently asked questions is available at: http://wiki.panotools.org/Hugin_FAQ To post to this group, send email to hugin-ptx@googlegroups.com To unsubscribe from this group, send email to hugin-ptx+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/hugin-ptx
