With 3-D movies such as Avatar and Toy Story 3 breaking box office records, 3-D televisions hitting store shelves, and 3-D computer gaming consoles due later this year, stereoscopic technology is hot.
Watching a three-dimensional presentation requires that each eye see the subject from a slightly different viewpoint. The spacing between our eyes produces that effect automatically for objects we see in real space, but seeing stereo from images on a flat screen requires slightly different left and right image pairs, with each image fed to only one eye.
For a single observer, that problem was solved more than 150 years ago, with dual-eyepiece viewers that presented each eye with one of the paired images. For groups of viewers it has been achieved by showing both images on a single screen, with audience members wearing eyeglasses that let each eye see only one of the images. Movie theaters project both images at once, through polarizing filters at right angles to each other, while the audience wears polarizing glasses. Today's 3-D televisions rapidly switch between left and right images while viewers wear glasses with synchronized shutters.
The glasses required for today's 3-D TVs and theater showings can be annoying and uncomfortable, however, to say nothing of expensive. Handheld autostereoscopic, or glasses-free, 3-D devices are already entering the market, but researchers are laboring to produce glasses-free 3-D for more than one or two simultaneous viewers.
The main challenge so far is to keep the left and right images separated (low crosstalk) without compromising the amount of projected image light reaching the viewers' eyes (high optical efficiency). The most common approaches—parallax barriers and lenticular lens screens that produce left-right images—have trouble maintaining left-right image separation without reducing image brightness and resolution for multiple concurrent viewers.
But a team of engineers from National Chiao Tung University, in Hsinchu, Taiwan, seems to have overcome that challenge in a glasses-free 3-D projection system. They described their work in "An Autostereoscopic 3-D Display System Based on Prism Patterned Projection Screen," in the March issue of the IEEE Journal of Display Technology.
With the currently common parallax barriers, which are opaque vertical bars that block different portions of the display for each eye, narrowing the slits between the bars reduces the likelihood of each eye's seeing the other's image (crosstalk) but also reduces optical efficiency by blocking more of the light coming from the images. Widening the slits does the opposite.
Different problems apply to the other common approach, the lenticular screen, covered with molded lenses that direct left and right images to the proper eye. Unlike parallax barriers, they don't reduce optical efficiency. But they require specially created images, have high crosstalk between left and right images, drastically reduce image resolution in multiple-viewer systems, and can appear blurred or disorienting except when viewed from a precisely defined, fixed distance and viewing angle.
What the researchers—Wallen Mphepö , IEEE Member Yi-Pai Huang, and IEEE Fellow Han-Ping D. Shieh—have developed is an autostereoscopic projection system using a single projector and a curved screen whose surface is covered with many tiny prisms. The microprisms' position, orientation, and angle must be carefully computed. The projection system uses interlaced left- and right-image pixels that strike the screen at different angles of incidence. The prism surfaces are angled to reflect each image's pixels only to the appropriate eye, and the angles vary with each prism's position on the screen.
"By designing the prisms to reflect the incident light from the projector to specific pre-computed left and right eye locations at the viewing distances," says Mphepö, the lead researcher, "we obtain a sweet spot where there is maximum separation of left and right pixels."
The screen is then parabolically curved in the horizontal plane to keep the reflected image rays within the viewing zone. In its current state, the system is applicable to front- or rear-projection screens, but not to conventional flat-screen televisions. Flat-screen capability is one of the research team's targets, though. Any commercial application requires further work.
"We think our current system could accommodate around 115 concurrent viewers, using a 32-foot screen," Mphepö says. "However, if we scale up to 65-foot and larger screens—as we eventually will after we finish our work with 32-foot screens—we expect to accommodate more than 600 concurrent 3-D viewers." Initially, presentations to large audiences will require unconventional seating arrangements with the viewers staggered in rows within a conical space, he says.
"The system is more cost-effective for large screen sizes because the surface machining tolerances for the microprism surfaces increase with increasing diagonal size," he says. "The smaller the screen the more stringent the prism surface parameters."
The team has made improvements to the system since its paper was published. With the system outlined in the paper, Mphepö says that if the viewers moved their heads too much they would see either a reversed 3-D image, where the pixels were sent to the wrong eyes, or just a 2-D image. In the newest version a viewer would have to move drastically—for example, to stand up and move to a new location—for the head to be out of the 3-D viewing zone. And the researchers have eliminated the reversed 3-D image problem.
The team also has been working on other issues such as designing cheaper fabrication processes and experimenting with alternative materials for the prismatic screen. The main obstacle, Mphepö says, is finding a company that could fabricate a large enough screen.
Working at their current pace with the resources at their disposal, the researchers estimate that the screen design could be ready to commercialize in three to five years. "But it could be a lot sooner if we get a suitable screen company to collaborate with or see an infusion of resources into the project," Mphepö says. "The producers of films like Avatar have had to invent new and radical 3-D image-capturing methods. We display researchers need to be just as radical and inventive, if not more."