On Wed, 5 Feb 2025, 03:45 Greg 'groggy' Lehey wrote: > > Thanks! But your reply poses more questions than it answers. >
Questions are good! > lot of parallax error in this scene. > > In this situation you should only use control points on the furthest > > features, and ideally only features that are near the likely stitch > > seam (automating this would be a useful Hugin feature). > > Does this mean that you set the control points manually? > Some manual control points were needed. The main thing was to delete control points on foreground objects, and background points away from the likely seam positions. > The photos are circular fisheye, so you need to check that there is > > a circular crop on each of the photos, > > How do I do that? You've seen what the images look like; should > anything else be done? > In the mask/crop tab, make sure there is a visible 'crop circle' for each image, there is also a checkbox used to link the crop for all images. > this is the cause of the 'fleur-de-lis' artefact. > > My guess is that you loaded the images differently. I've been doing > it via the fast panorama preview I didn't use the preview, but the functions are the same. Once you have this rig calibrated in the lens database, you should be able to do everything from the preview. Thanks. Yes, this works for me, but I still don't know how to repeat > it. I've tried using the main window, and I discover that Hugin > decides that it's a circular fisheye—is this heuristics? > Yes, Hugin stores lens settings and uses them next time. But it also changes the focal length from 4 mm to 2.903 mm, which > sounds like a bug. That means that if I manually enter the focal > length (from versions without Exif data, for example), it gets > incorrect information. At 4 mm the calculated angle is 185.92°, and > at 2.903 mm it's 256.18°. The stitch only uses the central 120 degrees of each photo. Control points in the periphery, ie. areas that are not contributing to the final image, are counter-productive (because of the large parallax error). So, in this scene, the angle of view could be wildly off and it wouldn't make any difference - in fact letting it go wild probably helped lining up the bits in the centre we do care about. When you do calibrate this lens, using four photos and rotating around the no-parallax point, you can ensure there are control points in the periphery, and you will get an accurate angle of view. Also, when you calibrate the lens, there are other input fisheye types that might be a better fit for this lens, try them and see. So: I think we have at least two bugs here, the fleur-de-lis pattern > for the second image in the fast panorama preview and the incorrect > angular calculation. > Hugin doesn't set a default crop circle when you pick 'circular fisheye'. Potentially it could guess this or even try and figure it out from the image itself, but for now you have to tell it where you want it. -- Bruno > -- 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 [email protected]. To view this discussion visit https://groups.google.com/d/msgid/hugin-ptx/CAJV99ZjNkCWk1ahf0qktVS%3Df%2B2zBvjW6HB-GCPEhUvEzV_rcOw%40mail.gmail.com.
