18 
Well, maybe not. But a job listing on Apple’s jobs portal is asking for a 3D expert to come onboard the mothership, so it clearly has a plan to incorporate three-dimensional tech of some sort in a future iOS device.
The job listing specifically asks for “a Computer Vision specialist” who’s experienced in “Multi-view stereo and 3D reconstruction” and the “Inpainting of occluded geometry and texture data” whatever that may mean. It sounds like there may be a gaming element to the position as well, with Apple also asking the potential applicant to know about “Cameras and surfaces in a 3D environment”.
Apple hasn’t yet launched any 3D products, leaving it to the likes of HTC and LG to test the water (rather unsuccessfully) in the mobile arena. [9to5Mac]
Image credit: Apple’s potential new HQ, via Cupertino.org
Sony Hires Woz as Advisor for Steve Jobs Biopic
Why Would Nokia Want to Hire an Android Expert? (Updated)
Samsung/Apple Legal Expert Not Strictly Hired by Samsung, But Still...
well they already filed a patent that hinted at 3D camera technology, so using my amazing logic powers i’m guessing that the two are connected. maybe they’ll have full, working AR in future products, working like it does on a 3DS
This sounds more like kinnect than holograms. Maybe the “Inpainting of occluded geometry and texture data” bit means that it will recreate you as an avatar inside the computer.
I Think that the “Inpainting of occluded geometry and texture data” bit is referring to up scaling 2D images to 3D. I’m fairly outnumbered on Gizmodo, I love 3D, it works a charm on my 3DS and if anyone can make it worthwhile on a mobile device, it’ll be Apple.
Ugh, I hope they give up, the last thing we need is Apple telling people 3D is a good thing. People will be stupid enough to believe them.
Re: inpainting of occluded geometry and texture data.
I guess there are two big issues related to this topic that need to be addressed in the context of multiview stereo. The first is that even without occlusion, regions of an object with sparse texture might need some additional regularization term to effectively help fill in their geometry. An example might be a large, homogeneously textured wall. It’s very difficult to reason about it’s geometry at a local scale using multiple images (where does one point match up in another image if the wall looks the same almost everywhere!), but if you consider the extent of the wall, you can eliminate some of the ambiguity. Maybe it’s easy to reason about visual correspondence around edges which then can be used to constrain the optimization. The same concept may be applied to partially occluded surfaces as long as you can correctly reason about the occlusion itself.
Inpainting of texture data is more straightforward. These usually involve some sort of patch sampling or texture synthesis based approaches. One example of a product that successfully uses such an approach is the content-aware fill in Photoshop.