More Lytro thoughts

[jducoeur]

dervishspin mentioned wanting to capture the "3d-ness" of a scene. This sparked some thoughts about the possible future directions of this new camera technology, which seem worth a top-post:

The neat thing here is the sense that they are really on to something. This first model feels like the original iPod: a bit oversimplified, going to be frustrating in some ways, needs a lot of tweaking and tuning and a whole bunch more power. But I have a strong suspicion that in a few years, once they've evolved both the hardware and software a bit, it's going to change the whole way one thinks about photography.

Specifically on the 3d-ness -- one obvious thing to try (and I'd be surprised if they're not experimenting with it already) is a 3d version of this camera. Two lenses, both recording the complete light field like this -- I suspect there's a whole 'nother doctoral thesis on how you can stitch them together, but I'd bet that you can eventually wind up with a fully immersive image, that looks truly 3d *and* allows you to refocus. It probably could even be tuned to your specific eyes, to work better for many people than most current 3d does.

Then for extra credit, take that idea and combine it with hardware that follows your pupils and refocuses the image based on your current focal distance. Once the software is fast enough to work in perceived realtime (probably 1-2 orders of magnitude faster than their current demonstrations), you can have a "picture" that is immersive beyond anything currently possible -- you can actually "look around" the image intuitively. Now imagine giving a tour through a panoramic 3d image with your eyes, with the image refocusing and zooming based on your eye movements as you show participants around what's in the image. From what I know of the technologies, that seems like it's plausible within five years or so.

I always love it when I get that little spine-tingling sense that a bit of science fiction is about to become real...

Comments

[talvinamarich.livejournal.com]

Hack one together with a Kinect:

http://www.youtube.com/watch?v=7QrnwoO1-8A

[laurion]

I think the image would be 3-D the same way stereoscopic images (the kind where you split your eyes into focusing slightly apart from each other) are 3-D. You'd be able to get some shifting of view, some parallax, but you wouldn't be able to get true shifting of the POV. More importantly thought, it could be used to recreate a scene with higher detail. Imagine being able to quickly and effortlessly generate a computer model of a room, with everything in exact placement. Home inventory, crime scene, archaeological recordings, etc. Imagine extending this to x-ray technology. Xrays are cheap compared to MRI or CAT, but are two dimensional at the moment.

[mindways]

I thought I'd read somewhere that you could extract 3D information / do a 3D rendering of a *single* Lytro photo.

*pokes around* Ah, here.

Two would probably be more complete, but I believe they're touting "can build 3D models from a camera shot!" as one of their potentials.