--- In PanoToolsNG@yahoogroups.com
, "alfred.molon" <alfred.molon@...> wrote:
> It can be done manually, by scaling the images appropriately, but could this feature (i.e. ability to merge photos with different focal lengths) be implemented in PTGUI?
> It's helpful when you find out after the shooting session that it would be a good idea to merge some photos not originally captured for a panorama. These photos often have been taken with different focal lengths. Apparently Autopano Pro can do this.
I don't know about PTGUI, but hugin does it well. You do not need to do anything manually - hugin reads the lens parameters from the EXIF data of the image (if they are included, if your images are from raw, the converter might not pass the EXIF data on, in which case you have to enter them manually).
The SIFT and SURF algorithms for automatic control point generation are fairly scale-insensitive - to give you an idea: if I take a shot with my EFS 18-55 mm at 18 mm and several at 55, there is no problem matching the shots even though they're about a factor of three apart. If the images are low-res, bas quality or the focal lengths are wider apart, things can become more difficult. I just recently dug into my old files and found a series of shots of the Nanda Devi Range taken from 80 km away with an Ixus 30, wide angle, zoommed 3.2 and through the ocular of my pair of binoculars. These three levels of magnification would not match automatically, but the wider angle shots were great to pin the tele shots onto and fill in gaps, and all it really took was manually setting a few control points, which then would even fine-tune, once they'd been positioned about right, in about 50% of the cases.
Needless to say you should have no parallactic errors int he image series (or live with the artifacts, or retouch them later.) When optimizing, after you have the images pretty much in place, optimize 'everything' - this will adapt the different lens correction parameters of the different focal lengths probably better than preprocessing the images with PTLens or equivalent.
Again, I don't know about PTGUI, but I reckon that there as in hugin you can either have the images stitched by the GUI or hand-stitch them yourself. Hugin, if asked to stitch images of different focal length, will do so quite happily and you end up with a blown-up version of the wide-angle data, the other data considered 'redundant'. At least I have found no way to get it to not do so. If you have covered your subject fully in all resolutions, you can just throw out the wider-angle-shots before stitching - you might want to do such a thing if you use the wide-angle shots as reference and have pinned the tele shots onto them. If you also stitch the same project with just the wide-angle shots, you end up with two compatible images which you can hand-blend in the gimp.
An interesting alternative I've used with good results is to have the nona-ed images output and use enfuse on the command line like
c:\programme\hugin\bin\enfuse -o hand_fused.tif --wExposure=0 --wSaturation=0 --wContrast=1 --HardMask seqence*.tif
... so I fuse them as if they were a focus stack. The wider angle shots, nona-ed to the appropriat size, normally have less local contrast than the tele shots and get picked only where the tele data are bad (out of focus, or missing), and so you end up with an image with as many details as possible with gaps reasonably filled. For this technique to work, the images should align very well, otherwise ugly artifacts (double images etc.) appear.
Hope all of this is helpful. Happy stitching