Do you have an idea for an awesome feature we should add… or hate the way we’re currently doing something? Share your ideas and suggestions here.
Since Photosynth seems to have a really good grasp on how shots relate to one another, would it be possible for Photosynth to flag shots that would work well as stereo pairs? Even better, if it could handle any necessary geometric transformations and image cropping, you could potentially go through a synth with a stereo view of the scene.
(I realize I'm reaching at this point, but I figure if I don't ask I can't get it.)
That's actually a pretty reasonable idea.
Given that we have 3D points (the point cloud), I'd venture to say that any pair of neighboring images in one of the better synths would work well as a stereo pair (assuming there was enough parallax).
As for getting an actual stereo view, some of us have discussed it, but the main issue is that we currently project the images on single planes. In order to get good stereo you'd essentially need a depth-per-point approach. This is in the works, but not quite there yet.
Dreamy idea Tom. Good to see there are some visionaries who can grasp the current cutting edge and be "possibility thinking" the next possible steps. I've always been fascinated by 3D (considering its standard in real life). I've always wondered how previous 3D viewing technologies have arrived and then faded away (like the old Atari ST's LCD shutter glasses). I thought for sure we'd be seeing IR outputs on our cable/satellite TV boxes (or at least computers) for synching Left/Right shutter on a 3D TV channel by now. (then again, I've been trying for a while to get MS to add spell check (common sense?) to Messenger even if tapping into the installed MS Word dictionary). Sorry for the A.D.D. moment... Are you currently doing anything with stereo pairs? I've been dinking around with cross-eyed static 3D and considering creating a simple animated version of something with limited background movement but cycling subject movement like the Lynden windmill.
To answer your question, yes there are people doing work on stereo pairs, but I don't think you'll see any benefit from this work just yet. That said, doing something like finding two images taken at the right distance and putting up selective red-green pictures doesn't seem like a hard demo to create.
It might be a start to do as the OP suggested - flag up suitable pair candidates - and maybe present them as a separate 'cross eye' pair for viewing by those of us with the nack and immunity to eye strain headaches. No special equipment required - just a bit of practise. Going through a whole synth in 3D would be amazing but the resulting view might be rather small. I think you'd have to shoot it with 3D in mind - or the parallax would vary from view to view. Now that WOULD cause a headache - if not motion sickness
Why not simply flag image pairs with base-depth ratios near that of human perspective, and then apply epi-polar resampling to a parrallel image plane and kick out an red-green/blue image for the qualified pairs?