I am not saying that non-NPP issues are a shortcomunig of Hugin. I am saying, that morphing the controlpoints in a way a texture is mapped on a wireframe model (using baricentric coordinates in a triangle) will allow stitching non-NPP panoramas without visible seams. The biggest issue then will be occlusion, which will cause half-transparent areas. A workaround for this could be occlusion detection which should be possible by increasing the density of controlpoints in areas that are (partially) not mappable.
Am Freitag, 22. April 2016 14:12:32 UTC+2 schrieb Sean Greenslade: > > On Thu, Apr 21, 2016 at 03:14:21PM -0700, Simon Bethke wrote: > > Hi, > > I am an amateur photographer and lately got a Gear VR device so I wanted > to > > try some VR Stuff. > > Testing free panorama tools, I found that Hugin was the only free tool > > providing decent results. Also, my 'Fisheye' lens uses Stereographic > > projection, so most other tools are not even able to stitch my input > > material correctly. > > > > Now, the last days, weeks, ... even months(?) ... I am trying to find a > > good way to capture stereoscopic panoramas with my one camera. After > alot > > of trial and error, I found that the best method for me is to place the > > camera left and right the NPP and take 12 images for the 360° circle. > Now > > with hugin, I align them and export the frames but don't stitch them > yet. > > For that I use the multiblend <http://horman.net/multiblend/> tool and > > first extract its seams. Then I edit the seams that left and right eye > > matches and blend with these optimized seams. My problem with this > method > > is, that by taking 12 left-eye images and afterwards 12 right-eye > images, > > the time between the left and right eye is to long: Clouds pass by and > even > > the sun travels. Flipping between left and right is inpractical and > causes > > many mistakes if I try to do it fast. > > > > What I would like to do is to have the camera in front (not left or > right) > > of the NPP, take 12 photos and use only right image-halfes for the left > eye > > and vice versa. This works more or less, but here the big issue is, that > > the paralaxe gets too strong and the stiching result is a mess with alot > of > > visible seams. At this place I wish it was possible, that hugin looks > for > > areas with bad control-points. If the photos were taken from a tripod > (with > > paralaxe), the error should equal the depth of the control points. At > least > > if there are areas with a bigger error (and similar error vectors) and > > others with less error, the software could treat this as depth. Now with > > depth information of the control-points, this becomes something similar > to > > visual sfm. The controlpoints are like a sparse point cloud and I'd > suggest > > simply interpolating the areas between the controlpoints to get a > perfect > > stitching result. > > > > If hugin could actually support stitching photos that are taken off the > > NPP. It could also assist creating stereoscopic panoramas. > > > > What do you think about this? Is Hugin still actively developed or is it > > rather a complete product which is just supported? > > > > Kind Regards, > > Simon > > It's not a shortcoming of Hugin that non-NPP pano sets are hard to > stitch. It's simply the way optics work. There will be parallax errors > if you're not spinning around the NPP, and there's nothing that can be > directly done about that. Parallax errors are extremely difficult for > software to handle automatically. You can of course use Hugin to stitch > non-NPP sets, and even handheld shots. Often times, some clever manual > masking can reduce or even eliminate the more severe parallax errors, > but that relies on there being "fudgable areas," e.g. blank white walls > where you can hide the misaligned seam. > > The advice from that forum post in the other reply seems to be pretty > good. Keep the number of photos as low as possible to keep the number of > seams down, and avoid very close objects to the camera (which cause the > most severe parallax). > > The problem is that 3D and pano-stitching are essntially at odds with > each other. 3D uses parallax to imply depth, while panos want as little > parallax as possible (ideally zero), since it causes stitching artifacts. > > --Sean > > -- 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 on the web visit https://groups.google.com/d/msgid/hugin-ptx/28173164-4d97-4558-b700-8792ce2821cc%40googlegroups.com. For more options, visit https://groups.google.com/d/optout.
