37.1 Stereo glasses🔗⧉

You have two eyes, set about six and a half centimeters apart, and they see the world from two slightly different positions. Hold a finger close and wink each eye in turn: the finger jumps sideways against the background. That jump is binocular disparity, the difference between the left and right retinal images, and the visual system fuses the two views into a single percept with vivid depth, a process called stereopsis (Chapter 5.8, Photographic representation and its limits). Disparity is one of the strongest depth cues we have, and it is exactly the cue a flat photograph destroys: a single picture, shown to both eyes, provides binocular cues to the flat display surface, while pictorial cues can still produce a strong impression of scene depth.

The oldest fix for flatness attacks this directly. If the flatness comes from showing both eyes the same image, then show each eye a different image, specifically the image that eye would have seen from its own position. Build a pair of pictures shot (or computed) from two viewpoints a human eye-separation apart, contrive some optical trick so the left picture reaches only the left eye and the right picture only the right eye, and the brain does the rest: it fuses them and perceives depth. That is the whole idea of stereo, and it predates photography. Everything in this chapter is a different gadget for the same trick.

37.1.1 Wheatstone's stereoscope (1838)🔗⧉

Charles Wheatstone built the first stereoscope in 1838 Wheatstone 1838, a year before the daguerreotype was announced, so the earliest stereo pairs were hand-drawn perspective sketches, not photographs. Wheatstone's insight was that depth from two eyes is not some mystical faculty but a direct consequence of the two images being geometrically different, and he proved it with an instrument that fed each eye its own picture. His design is the mirror stereoscope: two mirrors set back-to-back at the bridge of the nose, angled at forty-five degrees, each reflecting a picture mounted off to one side into the corresponding eye. Look in, and the left eye sees only the left drawing, the right eye only the right; the brain fuses them into a solid scene floating in front of you. It was a laboratory demonstration that disparity alone produces depth, and it remains the cleanest statement of what every device in this chapter does.

David Brewster made the instrument practical by replacing the mirrors with a pair of prismatic lenses Brewster 1856, one per eye, mounted in a small box over a side-by-side stereo card. The lenses both magnify and deflect, so the two halves of a single card fuse comfortably without the bulky mirror geometry. When photography arrived, the stereo card, two photographs of the same scene shot from eye-separated viewpoints, became cheap to mass-produce, and Oliver Wendell Holmes designed an open, handheld viewer (1861, deliberately unpatented) that drove the price down further. The result was a Victorian craze: by the 1860s the parlor stereoscope was a fixture of middle-class homes, and people collected stereo cards of foreign cities and natural wonders the way later generations collected postcards. The lineage from Wheatstone's mirrors to the Holmes viewer to the 3-D cinema glasses in your hand today is one unbroken idea: a different image to each eye, implemented in different devices.

37.1.2 Routing a different image to each eye🔗⧉

The stereoscope solves the routing problem mechanically: physical barriers and separate optical paths keep the two images apart. But 3-D cinema and 3-D television want one screen, seen by many viewers at once, with the two images superimposed on the same surface, so they need a way to multiplex the left and right images onto one display and then de-multiplex them at the glasses, sending each eye only its own channel. Four schemes dominate, and they differ in what dimension they use to keep the channels apart: color, polarization, time, or wavelength.

Anaglyph separates by color. The left image is printed in red, the right in cyan (blue-green), the two superimposed; the glasses carry a red filter over one eye and a cyan filter over the other. The red filter passes the red (left) image and blocks the cyan one; the cyan filter does the reverse. It is cheap: the glasses are cardboard and colored plastic, and the image prints on ordinary paper or any screen, which is why it dominated comic books, cereal boxes, and some 3-D films. Much of the 1950s theatrical boom used polarized projection rather than anaglyph. The cost is color: you have spent the color channels on routing, so anaglyph images are badly color-degraded and prone to ghosting (each eye faintly sees the other's image where the filters leak).

Polarization separates by the orientation of the light wave. The projector puts the left and right images out in two orthogonal polarization states, historically linear (one image polarized vertically, the other horizontally) and now usually circular (left- and right-handed, so the separation survives a viewer tilting their head). The glasses are passive polarizing filters, each passing one state and blocking the other. Because polarization is independent of color, the images keep full color, and the glasses are cheap and unpowered, which is why polarization is the workhorse of commercial 3-D cinema. RealD, the most widely deployed system, uses a single projector with a fast electro-optical modulator that flips the circular polarization in sync with alternating left/right frames; IMAX 3-D classically used two separate projectors with orthogonal polarizers. The one catch is the screen: polarization-preserving "silver" screens are required, because an ordinary matte screen scrambles polarization on reflection.

Active shutter separates by time. The screen shows the left and right images in rapid alternation, and the glasses are battery-powered LCD shutters that go opaque and clear in lockstep: black out the right eye while the left image is up, then black out the left eye for the right image. Synchronized fast enough (each eye effectively sees its half-rate stream), the brain fuses a full-color stereo image. Active shutter keeps full color and resolution and works on an ordinary screen, but the glasses are heavier, need power and a sync signal, and the time-multiplexing can show flicker or crosstalk if the panel does not switch cleanly. This was the dominant scheme for the wave of 3-D televisions around 2010.

Wavelength multiplexing separates by spectrum. Dolby 3-D splits the visible spectrum into two interleaved sets of narrow red, green, and blue primaries, call them $R_1G_1B_1$ for the left eye and a slightly shifted $R_2G_2B_2$ for the right. The projector shows each eye's image in its own primary set; the glasses carry precision interference (dichroic) filters that pass one comb of wavelengths and block the other. Both eyes still get a full red-green-blue image (just built from slightly different primaries), so color and brightness are good and no special screen is needed, but the filter glasses are expensive optical parts, which is the scheme's trade-off. This color-comb idea is sometimes called spectral comb filtering or, after the company that pioneered it, the Infitec method.

37.1.3 Shooting and synthesizing a stereo pair🔗⧉

Having a way to deliver two images is half the problem; the other half is making a good pair. The two viewpoints must be separated horizontally, and the natural choice is the human interaxial (or interocular) distance of roughly 6.5 cm; shoot with two cameras that far apart, or render two virtual cameras that far apart, and the captured disparities approximate natural eye separation. Matching perceived scale also requires the intended display size, viewing distance, camera field of view, and viewer eye position. This faithful choice is sometimes called ortho-stereo. Stereographers deliberately depart from it for effect: a smaller baseline compresses the depth (useful for close subjects, or to keep disparities gentle), and a larger baseline, called hyperstereo, exaggerates depth, which is how you make a mountain range look like a tabletop model.

A second control is convergence: where the two cameras' optical axes cross sets the depth that will appear at the screen plane (zero disparity). Objects nearer than the convergence distance pop out of the screen (crossed disparity); objects farther recede behind it (uncrossed disparity). Convergence can be set by toeing the cameras inward or, more cleanly, by shifting the sensors/crop horizontally in post. The art of it is keeping disparities within a comfortable range: too much and the eyes cannot fuse the image (it doubles), or fusing it becomes exhausting.

And that exhaustion has a specific cause, the one this chapter hands forward to the rest of the part. When a virtual object floats one meter from the viewer, your two eyes converge as if fixating a point one meter away, so the vergence cue says "one meter." But the light actually comes from the screen, perhaps fifteen meters off, so each eye must accommodate (focus its lens) at fifteen meters to see the image sharply, so the focus cue says "fifteen meters." In the real world these two always agree: you focus and converge to the same distance, and the visual system has wired them together (a reflex called accommodative vergence). A stereo display pulls them apart. This is the vergence–accommodation conflict (VAC), and it is an important contributor to stereo discomfort, alongside excessive disparity, crosstalk, misalignment, and temporal artifacts, why filmmakers keep disparities modest, and why virtual objects cannot comfortably come too close. Fixing it requires a display that can put the focus cue at the right depth too. No stereo-glasses scheme can do that, because they all paint the image on one fixed plane. That is the frontier of Chapter 37.3, 3D displays with accommodation, and the reason this chapter is only the first rung of the ladder.