Autostereoscopy

Autostereoscopy is any method of showing a three-dimensional image with stereo depth without requiring the viewer to wear special glasses or other headgear. An autostereoscopic display uses optical elements placed over an ordinary flat panel, or other directional optics, to send a different image to each of the viewer's eyes, so that the viewer perceives depth. The two most common light-directing elements are the parallax barrier, a mask of fine slits, and the lenticular sheet, an array of small cylindrical lenses.

The technique is more than a century old: Frederic E. Ives patented a line-screen "parallax stereogram" in 1903. Electronic versions have appeared in handheld game consoles such as the Nintendo 3DS, in 3D monitors and laptops, and in light field telepresence systems such as Google's Project Starline. Autostereoscopic displays differ from head-mounted displays, which give each eye its own screen, and from glasses-based stereoscopic 3D, which puts two images on one screen and uses filters worn by the viewer to separate them.

Reviewed 27 September 2026. Checked the Dodgson 2005 and Holliman et al. 2011 survey claims, the Ives patent, Nintendo 3DS, Acer SpatialLabs, Samsung Odyssey 3D, Looking Glass, Project Starline, HP Dimension and Lanman-Luebke 2013 details against their cited sources. About review dates.

Definition

In his 2005 survey of the field, Neil Dodgson of the University of Cambridge Computer Laboratory lists four depth cues that are missing from ordinary 2D images: stereo parallax (a different image in each eye), movement parallax (different images as the head moves), accommodation (the focusing of the eye's lens) and convergence (both eyes turning toward the object of interest). All 3D display technologies provide at least stereo parallax. Autostereoscopic displays are the subset that does this "without the viewer needing to wear any special viewing gear".

Dodgson sets these displays against two other families. Glasses-based systems present both images in the same display plane and let the glasses decide which image each eye sees, using colored filters (anaglyph), polarizing filters or shutter glasses. Headsets mount one small display in front of each eye; Dodgson notes that they are limited by the need to wear the headset and by the isolation from the real world, and that see-through headsets always show their image against the real background. When combined with head tracking, both families can also give movement parallax for a single viewer.

How it works

Every autostereoscopic display uses some optical component to make different images visible on the same screen plane from different points of view. In the most common designs the slits or lenses are aligned with the columns of pixels on an LCD panel, so that one set of columns is visible from the left eye and another from the right eye. The result is a stereoscopic "sweet spot" at a designed viewing distance, usually with repeated viewing positions to either side.

Parallax barrier

A parallax barrier is a mask of opaque and transparent strips placed in front of the pixel raster, so that each eye sees light from only every second pixel column. The Nintendo 3DS used this approach: its screen, made by Sharp, added a second liquid crystal layer next to a conventional LCD and backlight, forming thin vertical strips that block part of the light and direct the rest alternately to the left and right eyes for a single viewer at a set distance, usually around 30 centimeters. Holliman and colleagues note that barriers have been less successful than lenticular sheets because the barrier blocks most of the light and makes the display considerably dimmer. IEEE Spectrum reported that the 3DS compensated by raising backlight brightness, which increases power draw.

Holliman and colleagues describe three extensions developed in the 2000s: a wavelength-selective filter array from 4D-Vision that gives red, green and blue their own barrier; a dynamic barrier that changes quickly enough for every pixel to reach the eye within the eye's integration time; and designs with two fully addressable panels, where the front panel acts as a computed barrier.

Lenticular lens

A lenticular display places an array of cylindrical lenslets over the pixels. Each lenslet directs light from adjacent pixel columns to different viewing slots at the ideal viewing distance. Lenticular sheets do not block light the way a barrier does, but with vertically aligned lenslets the display magnifies the gaps between pixels, producing dark zones between viewing slots, and horizontal resolution is divided among the views while vertical resolution is not. Holliman and colleagues give the example of a 1920 x 1080 panel that becomes a four-view display with 480 x 1080 pixels per view.

The standard fix is to slant the lenses. In 1996 Cees van Berkel at Philips produced a seven-view lenticular display with the lenticulars slanted relative to the pixels, a technique also proposed by Winnek in 1968. Slanting removes the dark zones and spreads the resolution loss across both the horizontal and vertical directions. Dodgson notes that Stereographics produced a range of nine-view displays using this design. Holliman and colleagues describe a two-view design by Harrold et al. that adds a liquid crystal switch and polarization-activated microlenses so the 3D effect can be turned on and off electronically, keeping comparable brightness in 2D and 3D modes; its crosstalk between the two channels was below one percent at the central viewing position.

Viewing zones and pseudoscopic images

A simple two-view barrier or lenticular display produces a repeating pattern of left-eye and right-eye zones in front of the screen. An eye inside one of these zones sees either the left or the right image; an eye between zones sees a mixture of both. Dodgson points out that even at the ideal viewing distance there is a 50 percent chance that the viewer's eyes fall in the wrong zones and see a pseudoscopic image, with depth reversed. The viewer must also stay fairly still, and moving much closer to or farther from the ideal distance greatly reduces the chance of seeing a correct image. These limitations lead to the two main refinements: tracking the viewer, or adding more views.

Head-tracked and eye-tracked displays

A head-tracked display shows only two views but measures where the viewer is and places the left and right images in the correct zones, which prevents pseudoscopic viewing. If the rendering also uses the head position, the display can simulate movement parallax; otherwise it gives stereo parallax only. Such systems usually serve one viewer at a time, and Dodgson identifies tracking lag and the physical movement of the viewing zones as their main engineering problems. Holliman and colleagues note that tracking only the center line of the face is often enough to steer the views, and that Sharp's micro-optic twin-LCD prototype tested both infrared detectors and video tracking.

Several consumer and professional products use camera-based face or eye tracking. The New Nintendo 3DS, launched in Japan in October 2014, tracked the user's face with its internal camera and an infrared LED and adjusted a dynamic parallax barrier to match; Nintendo European Research & Development describes the result, "Super-Stable 3D", as giving a significantly wider viewing range than the original 3DS, which required the user to stay centered on the device. Acer's SpatialLabs system, announced on 27 May 2021, pairs a stereo camera above the screen that tracks the user's head and eyes with a UHD panel topped by an optically bonded liquid crystal lenticular lens that switches between 2D and stereoscopic 3D. Samsung's 27-inch Odyssey 3D G90XF monitor, a 3840 x 2160, 165 Hz IPS LCD officially launched in April 2025 at US$1,999.99, likewise combines eye tracking with a lenticular lens.

Multiview displays

A multiview display divides the space in front of the screen into three or more viewing slots and shows a different image in each. Anyone whose two eyes fall anywhere inside the overall viewing zone sees a 3D image, several viewers can watch at once from their own points of view, and moving the head gives horizontal movement parallax, with small jumps between slots. No tracking is needed. The cost is that every view must be generated and displayed at all times, whether anyone is looking at it or not. Holliman and colleagues cite a 60-view experimental display from LG Display as an example of the resolution problem: each view had only 260 x 480 pixels.

Looking Glass Factory's desktop displays (including the Looking Glass Portrait) are multiview devices of this kind. The company states that its displays show up to 100 views of a scene across a view cone roughly 58 degrees wide, without glasses or eye tracking, so that several people can see the 3D image at once.

Integral imaging and light field displays

Replacing cylindrical lenslets with a two-dimensional array of spherical microlenses gives parallax in both the horizontal and vertical directions. This is integral imaging, first described as a photographic technique by Lippmann in 1908. Each microlens needs a whole image beneath it, so integral displays give up much more spatial resolution than lenticular displays for the same pixel count; Holliman and colleagues also list a limited viewing angle, very limited depth of field and the difficulty of making good enough microlens arrays as challenges.

The term light field display, borrowed from computer graphics and computational photography, is now widely used for 3D displays of this type. Holliman and colleagues point out that all displays reproduce some form of light field, and that the more views a display has, the closer its light field comes to that of a real scene.

Other approaches

Dodgson groups autostereoscopic technologies into three classes. Spatial multiplexing (barriers and lenticulars) divides one panel's resolution among the views. Multiprojector systems use one projector per view, shining onto a special screen such as a double lenticular sheet; these are expensive and need precise alignment, but experimental systems with more than 100 views have been built. Time-sequential systems use one very fast display and an optical element that sends each frame to a different zone; the Cambridge display used a high-speed CRT with ferroelectric liquid crystal shutters inside the projection lens. Hybrid versions of the Cambridge display reached 28 views at 25 inches and 15 views at 50 inches.

Approach	Light-directing element	Typical views	Main drawbacks
Two-view spatial multiplex	Parallax barrier or lenticular sheet over an LCD	2	Half horizontal resolution per eye; fixed sweet spot; 50 percent chance of a pseudoscopic view
Head- or eye-tracked two-view	Barrier or lenslets with views steered to the tracked eyes	2	Usually one viewer; tracking lag
Multiview lenticular (slanted)	Slanted cylindrical lenslets	Commercial displays typically 7 to 9	Resolution divided among views
Integral imaging	2D array of spherical microlenses	Horizontal and vertical parallax	Large loss of spatial resolution; narrow viewing angle; shallow depth of field
Multiprojector	One projector per view onto a special screen	More than 100 in experimental systems	Cost of one projector per view; alignment
Time-sequential	One high-speed display plus directional optics	15 to 28 in the Cambridge display	Needs very fast display devices

History

The parallax barrier predates electronic displays. Frederic E. Ives of Philadelphia filed a patent on 25 September 1902, granted on 14 April 1903 as US patent 725,567, for a "parallax stereogram": a photograph containing two interlaced, laterally displaced images, viewed through a screen of opaque lines with clear spaces between them. At the right distance each eye sees only its own image. In 1908 Lippmann described integral photography, the basis of later integral imaging displays. Dodgson notes that researchers had been making two-view displays with parallax barriers or lenticular sheets for more than a century by 2005.

Spatially multiplexed displays, which put a barrier or lenslets over the screen, depend on flat panels: the optics must line up with a fixed pixel pitch, and Dodgson notes that building a CRT with a precise enough pitch is extremely difficult, which in practice requires liquid crystal or plasma panels. In the mid-1990s, 40-view and 72-view experimental displays combining spatial multiplexing with multiple projectors were reported, and van Berkel's 1996 slanted-lenticular display at Philips made multiview lenticular panels with usable resolution possible. Writing in 2005, Dodgson said that usable glasses-free systems had been available for a few years and had found uses where depth perception is vital (scientific and medical visualization, remote manipulation of robots in dangerous environments) or where the novelty sells (games and advertising). He questioned whether they would move beyond these niches, and observed that the extra cues help only at ranges of a few meters, so they add nothing to applications such as flight simulators.

The Nintendo 3DS handheld game console, launched in Japan in February 2011, brought an autostereoscopic screen to a mass-market game device. Its top screen had 800 x 240 pixels, giving each eye a 400 x 240 view, from a Sharp-made parallax barrier panel. Face tracking followed in the New Nintendo 3DS in 2014. In the 2020s, eye-tracked lenticular displays reached laptops and monitors through Acer SpatialLabs (2021) and the Samsung Odyssey 3D (2025).

Applications in VR and AR

Autostereoscopic displays are an alternative to headsets for showing stereoscopic 3D content, and several products position them that way. The clearest case is telepresence. Google announced Project Starline, an experimental 3D video-calling booth, at Google I/O in May 2021, describing its screen as a light field display that shows the remote person with depth and without glasses or headsets. The system paper, published in ACM Transactions on Graphics in 2021, describes it as a "head-tracked autostereoscopic display" that uses a lenticular display to provide stereopsis and motion parallax without wearable equipment, alongside 3D capture, compressed color and depth streams and spatial audio. Google has since commercialized the technology as Google Beam. The first Beam device, HP's Dimension, uses a 65-inch light field display and was priced at US$24,999, with Beam licenses sold separately; it went to select customers from late 2025 and, as of September 2026, is available to customers in the United States, Canada, the United Kingdom, France, Germany and Japan.

Desktop autostereoscopic screens are also used to view and edit 3D content without putting on a headset. Acer aimed SpatialLabs at creators working with 3D models, with software for viewing models and a developer program for Unreal Engine. Samsung supplies 3D content for the Odyssey 3D through its Reality Hub app and offers AI conversion of standard video to 3D. Looking Glass Factory's software turns 3D scenes from engines such as Unity into the multi-view "quilt" images its displays need.

The optics of autostereoscopy have also been carried into head-mounted displays. In 2013, Douglas Lanman and David Luebke of NVIDIA demonstrated a near-eye light field display: OLED microdisplays from a Sony HMZ-T1 viewer covered with microlens arrays in place of the usual magnifying eyepieces. The authors describe the design as sharing similarities with integral imaging displays, and report that each modified eyepiece was 42 x 31 x 10 mm with a 0.7 gram microlens array, compared with 43 x 31 x 37 mm and a 57.7 gram lens for the stock eyepiece, at an estimated 146 x 78 pixels and a 29 x 16 degree field of view. They propose it as a way to address the accommodation-convergence conflict that occurs with existing stereoscopic displays, at the cost of reduced spatial resolution.

Limitations

The problems described by Dodgson and by Holliman and colleagues apply to most current designs:

• Resolution. Spatially multiplexed displays divide the panel's pixels among the views, so each eye or view gets only part of the native resolution.
• Brightness. Parallax barriers block much of the backlight, and the loss grows as more views are added.
• Viewing position. Untracked two-view displays work only in fixed zones at a set distance; tracked displays generally serve one viewer. IEEE Spectrum reported that on a barrier display such as the 3DS, moving away from a head-on position can let the left eye see the image meant for the right eye, and can cause ghosting.
• Crosstalk. Light from one view leaking into the other eye produces ghost images; Holliman and colleagues state that low crosstalk is very important for high-contrast images with significant depth.

See also

• Stereoscopy
• Stereoscopic rendering
• Light field display
• Integral imaging
• Parallax
• Binocular vision
• Holographic display
• Vergence-accommodation conflict