Revision as of 17:15, 1 November 2011

3D Display Technologies

It was the introduction of consumer-level (aka COTS) 3D Displays in 2006 that was the primary catalyst that made-reasonable the ability to construct a VR display affordable to small research efforts.

Since the introduction in 2006 of Active-stereo DLP TVs from Samsung and Mitsubishi, there have been a multitude of products released, using a wide variety of technologies. Not all of these consumer-3D technologies work well as VR displays however. Thus here we will sort out the advantages, disadvantages and disqualificating features of each.

Features to evaluate include active vs. passive stereo, auto-stereo, DLP vs. LCD vs. plasma technologies, and more.

Method of Stereoscopic presentation

The two primary methods of presenting separate views to each eye (Stereoscopy) are frequently referred to as Active and Passive stereo. These names refer to the need (or not) for active electronics contained within the eye-glasses used to separate the left and right views. A third technology uses thin barriers placed at the screen to direct the left and right views in particular directions, allowing users to position their eyes such that no glasses are required for the stereoscopic pair. This glasses-free technology is referred to as auto-stereo displays. A fourth method of presenting stereo that certainly falls into the realm of low-cost is the color-filter based method referred to as Anaglyphic stereo. Anaglyphic stereo is still useful in a pinch, but is not recommended as a technology for daily 3D viewing.

Active Stereo

Systems using active stereo glasses have long been the mainstay of the virtual reality (and indeed scientific visualization) community. For many years, the company StereoGraphics sold a line of glasses known as "CrystalEyes". These glasses were used in CAVE-style VR systems throughout the world. There are now many manufacturers of 3D Active viewing glasses, including many of the television manufacturers.

Pros of Active Stereo

Higher frequency displays may render full-resolution

Cons of Active Stereo

Systems using infrared (IR) to transmit signal to glasses can interfere with IR of tracking systems and remote controls
More expensive
More fragile
Require battery replacement (or recharging)

Passive Stereo

Passive stereo systems have become the standard for theatre 3D presentations, and are growing in use for home 3D systems. Passive stereo works by multiplexing a stereoscopic pair of images through some form of light filter. The glasses then use simple optical filters to allow only the image for the desired eye to pass through the view. Light polarization is the technique typically employed by consumer 3DTVs. The other common form is the use of color filters. Color filters can range from the simple red/blue (anaglyphic) style to the more sophisticated Infitec multi-hue method.

Because polarized-light is used in most public theater 3D systems, passive 3DTV's are frequently marketed as "Cinema-style" 3D. However, the method of transmitting polarized-light passive stereo differs between theater and home. In the theater, the left & right images are shown simultaneously, allowing each image to be presented in full resolution. On a 3DTV, a technology called Film-type Pattern Retarder (FPR) is used, which dedicates alternating rows (or sometimes columns) to one eye or the other — thus reducing the physical resolution by half. (However, some experiments on perception suggest that the existence of redundant information in the two images leads to a result that is not as poor as 50% resolution.)

Pros of Passive Stereo

Less expensive to purchase (can even use glasses obtained attending a 3D movie)
Less expensive to maintain (don't break as often and no batteries)
Do not interfere with other IR-based technologies (e.g. tracking and remote controls)

Cons of Passive Stereo

FPR technology reduces resolution
NOTE: However, not all active stereo provides full-resolution either

Method of Stereoscopic Rendering

Rendering (as opposed to displaying) is how the 3D image is created. This image is then transmitted to the screen. There are a handful of techniques used to relay this 3D information to the screen. In some cases, the method of rendering will correspond with the method of display. However, there are also many cases where the display electronics will process the input pro

Time-interlaced

Time-interlaced, or frame-sequential, rendering works by sending complete images for the left and right eyes in a repeating, alternating sequence. (NOTE: depending on the display technology, this may be presented as time-interlaced, or converted into a spatial interleaving — such as used by FPR passive stereo displays.)

In OpenGL, the method of specifying time-interlaced rendering is to use the left and right frame-buffers. Enabling "quad-buffer stereo" will then set the graphics cards to output the two buffers in alternating succession.

Side-by-side (or tom/bottom)

A simple method to render and transmit stereo-pair images is to sqeeze them horizontally (or vertically), and merge the two half-sized images into a single full-width (or height) image with the left frame on the left and the right frame on the right (or the left on top, right on bottom).

These are easy to create in OpenGL (or just about any image manipulation package). The display then stretches the images appropriately, and renders them either time-sequentially, or through a passive filter.

Issue: Note that because the images are initially squeezed, half of the physical resolution is lost prior to transmission.

Issue: Note also that if the side-by-side technique is used in conjunction with FPR passive displays that reduce resolution vertically, the net effect is to reduce the physical resolution by one quarter.

Checkerboard or Alternating rows/columns

Another rendering method that merges left and right image-pair data into a single images is to interleave the data by row, column, or both (ie. checkerboard). Checkerboard format is a feature in DLP projection displays that has since been adopted by other displays for compatibility. As with side-by-side techniques, half of the physical resolution of each image is lost in order to merge them into a single image of the same size.

Rendering interleaved data in OpenGL is mostly straightforward, although one should be wary of the fact that OpenGL stencil buffers are not in effect with calls to glClear().

HDMI 1.4 (left/delta)

...