Thursday, October 2nd, 2014

The state of stereoscopic display

0

 January 18, 2011 SPIE Electronic Imaging, Burlingame, CA—Michael Robinson, chief scientist at RealD described the current state of stereoscopic 3-D in his keynote talk. Overall, the displays are more than adequate for the job, but the content capture needs to be tailored to the display, obviously a challenge when considering all possible display types. In addition, there are rendering issues for some applications like games.

3-D images have been around for a long time. In the Victorian era, stereoscopes provided 3-D depth cues by using displaced images for each eye. This technology closely approximates the actual viewing mechanism that people use. By the mid-20th century, these stereoscopes morphed into Viewmaster toys with images mounted on a disk.

In the '50's, the first generation of 3-D films appeared. They quickly faded away because the film makers overused the 3-D effects. In '80, Stereographics made a 3-D system using active shutter glasses. In 2003, its successor company RealD started working on projection equipment for movies. At this time, cinemas were starting to fight HDTV for eyeballs in the home, the first digital projectors appeared, and "Chicken Little" appeared on 2 screens in '05. Now the number of films and theaters with 3-D capabilities is growing to the thousands.

The majority of systems use a DLP-based projector and a special silvered screen that preserves the polarization. The system uses 3 flashes per frame and a LCD switching polarizer to project two images into a beam splitter and into separate quarter-wave polarizers to present left-handed and right-handed circular polarized images. The viewer looks through circular polarized lenses to get a left and right image. The latest projectors are now wide angle systems with a 1:1 width to throw ratio. This new machine allows smaller theaters to show 3-D films.

Other systems come from Sony and use liquid crystal on silicon (LCOS) instead the DLP. The 3-D image is comprised of separate top and bottom images that go through circular polarizers to a dual image lens which also adjusts for mixed green-magenta crosstalk.

Master Images uses a rotating wheel to insert the different polarizers. One problem is that tilting the head causes misalignment and ghosting. Dolby uses dichotic filters with one set of RGB filters using a shorter set of wavelengths and the other set using slightly longer wavelengths. This rotating filter is like the older anaglyph solutions without the severe color unbalancing. One issue is the cost of the matching dichotic filters for the glasses.

TV is mostly using active shutter glasses, because there is no change in the bill of materials. These systems need fast displays (> 200 Hz) and fast switching active shutter glasses. The systems project interleaved left and right images in sync with the glasses. Straight-on viewing provides high contrast, but there are many second order effects and the attenuation in the glasses is about 30 percent. Some systems use passive glasses with the circular polarizers built into the display.

Plasma displays use a more complex frame sequence. The pixel emits, is extinguished, re-addressed, and energized. The set needs to do this set of operations for every color and for every bit plane. The plasma panes needs a very high sub-field scan (600 Hz) and special high-speed phosphors that have high intensity but can turn off quickly. The active shutters are open for the duration of a field.

LCDs are already polarized but the issue is the switching speed per pixel. Nvidia extended the off cycles of the shutters in the glasses, to allow the LCD in the display to fade out. The passive displays send odd and even lines through a retarder. Persistence of the eyes is used to fill in the spaces. For monochrome displays, there are obvious low resolution effects.

Autostereo is mostly for handheld displays. One display technique it to use a parallax barrier that switches on and off. This hides some pixels in vertical stripes and the barrier can be turned completely off for 2-D displays. This type of display can get full display resolution. Other encoding formats include checkerboard patterns.

Lenticular screens can sit on top of a conventional display. The system need a very high resolution display as each image is only half resolution. Tilting the lens can help to hide the black spaces caused by off angle light. The rotation helps to fill in the gaps in the light. One problem with this solution, is that the display changes with viewing angle and has some intensity variations and moiré fringing issues.

Image crosstalk is an issue when trying to get lots of images to increase viewing angles. The number of images reduces the resolution by the inverse of the number of images.
Lenses are better than switchable barriers because they let more light through.

Directed illumination is another technique. 3M makes special light guides but rotation collapses the3-D effects into 2-D.

Capture for 3-D is difficult. The lenses need to be closely matched and aligned. One issue is the difference between capture and seeing. Capture equipment has a difference between the separation of the lenses and the separation of the images on the film or imager. Camera separation gives the scene depth and cannot be corrected easily. Film or image offset can be adjusted in post production.

Camera separation causes distortion of close features, and the effect is greater on large screens. To be comfortable, the zoom of the capture has to be the same as the zoom of the view and camera separation has to be equal to the eye separation. These requirements are not simultaneously possible, resulting in distorted 2-D images of a 3-D world.

To be acceptable, filmmakers must control perspective, roundness, and size and they can minimize these effects on the large screen. For perspective, an object that is half the size should be twice the distance away. For roundness, a head depth should be about the same as its height. And for size, a closer object needs to be smaller than one farther away.

For the smaller screens, capture is different. The 3-D issues can be adjusted with compositing. For gaming, the developer can pre-distort the images.

Creators can shift the depth scaling functions and can linearize the perceived depth cues. Tools can help to force the gradients to equal a fixed percentage of any point to provide uniform roundness. Mapping perspective gives relative motion in 3-D. If the user implements these corrections at the engine level, the corrections can be performed in real time and are applicable to any computer generated images.

Speak Your Mind

Tell us what you're thinking...
and oh, if you want a pic to show with your comment, go get a gravatar!