Thursday, March 21, 2013

One Day My Autostereo Will Come

Autostereo is the odd term for being able to see stereo without glasses. It may sound like the sound system for your car, but its not.  Lenticular images are an example of one form of autostereo.

I first became aware of how practical autostereo was at Ken Perlin's lab in 2000. There I learned the following things as we wrangled the demo for his autostereo paper to SIGGRAPH that year, some of them are technical and some are "industrial".    The point of this is that the technical issues are dwarfed by the industrial issues.

Is this the future of autostereo?

Here are two obvious technical issues followed by two, much harder issues involving commerce.

1. Most autostereo does not pay attention to where the observer is, and has just built into it a number of views that are visible from different angles. Alternatively, you can track where the user's head is in 3D and regenerate the two views that are necessary for that observer in that location which is what we were doing. But then you discover that most head and eye tracking algorithms are 2D not 3D. So reliable 3D head tracking became one of the hardest parts of the problem.

2. But if you are doing more than a store window display, then you will need to be able to generate 3D images from whatever the user sees, whether that be text, the OS window system, or some exciting 3D database of a giant robot. That requires good integration into a high performance (or a suitable performance) graphics device. This of course is very possible today, even on a handheld device (up to a point).

3. But then you discover that there are in fact very few manufacturers of the displays that go into devices. In fact, there are, or were, exactly 1 manufacturer left of CRTs and 4 manufacturers of LCDs and all the people who sell displays buy from these 5 manufacturers. And you discover that they can not just whip out 5 or 50 of something special, that is hideously expensive. And building a new factory is not less than billions of dollars, many billions.

4. On top of that you discover that the primary industrial uses of stereo in industry, are actually quite pleased with the quality and price of their LCD goggles. So that undercuts any productization you might consider that does not go to the consumer.

The point is that everytime you see a press release of a new cool technology or display, you should realize that almost exactly zero of these will reach the consumer. That is a little negative, but it has to do with the costs and risks of ramping up to the scale that would make it worthwhile in that very competitive market.

So we take all announcements of new technology displays, say with 6 phosphors instead of 3, or new autostereo with the grim realization that the probability that any of these becoming available at prices that anyone but the DOD can afford is nearly zero.

On that positive note, HP has announced what seems like a very cool way to autostereo on a handheld. It was just published in Nature.

You can read about it here.

The citation at Nature is at:

A multi-directional backlight for a wide-angle, glasses-free three-dimensional display

David Fattal,
Zhen Peng,
Tho Tran,
Sonny Vo,
Marco Fiorentino,
Jim Brug
Raymond G. Beausoleil

Multiview three-dimensional (3D) displays can project the correct perspectives of a 3D image in many spatial directions simultaneously1, 2, 3, 4. They provide a 3D stereoscopic experience to many viewers at the same time with full motion parallax and do not require special glasses or eye tracking. None of the leading multiview 3D solutions is particularly well suited to mobile devices (watches, mobile phones or tablets), which require the combination of a thin, portable form factor, a high spatial resolution and a wide full-parallax view zone (for short viewing distance from potentially steep angles). Here we introduce a multi-directional diffractive backlight technology that permits the rendering of high-resolution, full-parallax 3D images in a very wide view zone (up to 180 degrees in principle) at an observation distance of up to a metre. The key to our design is a guided-wave illumination technique based on light-emitting diodes that produces wide-angle multiview images in colour from a thin planar transparent lightguide. Pixels associated with different views or colours are spatially multiplexed and can be independently addressed and modulated at video rate using an external shutter plane. To illustrate the capabilities of this technology, we use simple ink masks or a high-resolution commercial liquid-crystal display unit to demonstrate passive and active (30 frames per second) modulation of a 64-view backlight, producing 3D images with a spatial resolution of 88 pixels per inch and full-motion parallax in an unprecedented view zone of 90 degrees. We also present several transparent hand-held prototypes showing animated sequences of up to six different 200-view images at a resolution of 127 pixels per inch.

No comments:

Post a Comment