The Illusion of Reality in Stereoscopy.

Fig. 1. Advertisement for Audioscopiks, 3D red/blue anaglyph short subjects, 1936.

We often hear the claim that stereoscopic 3D photography reproduces reality. That is seldom true. At best, 3D is a reproduction of reality, usually miniaturized. Likewise, motion pictures are only an artificial reality, for they capture, from just one point of view, a series of still pictures every 1/24 second and then present them on a flat screen, fooling the eye into thinking that objects in the scene are in continuous motion.

Stereoscopy captures scenes from two points of view, ideally separated by 63.5 mm (about 2.5 inches), the average separation of human eyes. These two pictures (a stereo pair) contain all the information needed to recreate the depth of the original scene, but they must be presented simultaneously to your eyes, properly sized and aligned. Various methods have been designed for viewing a stereo pair, with viewing devices (stereoscopes), by color filtration (anaglypics), lenticular screens, by projection onto a large screen, and even by learning eye gymastics to combine (fuse) the pictures of a stereo pair presented side by side. When viewing stereo pictures you can move your head, but you still only see the scenes from the original point of view of the camera lenses. (There are autostereoscopic viewing systems that capture scenes using a multi-lensed camera from many points of view and allow you to move your head when viewing stereo scenes, to select different points of view, but these are specialized and expensive.)

In spite of these limitations, when you view a stereoscopic picture, are you seeing at least a snapshot of the scene reproduced faithfully? Hardly, for several reasons.

Depth cues.

The stereoscopic disparity of the two pictures of a 3D pair is a strong indicator of distance, and is called "stereo parallax". Two eye-brain processes (called stereopsis) are involved. At close distances, of a few feet, the brain senses the convergence of the eyes due to nerves in the eye muscles. At larger distances, up to about 50 feet, the brain senses the difference betwen images on the retina (eyeball convergence plays little or no role at these distances). But there are other cues as well. Most of these don't function beyond about 50 feet.

• Angular size of familiar objects.
• Classical perspective, especially that due to straight lines converging with distance.
• Desaturation of colors and softening of clarity with distance (aerial perspective).
• Shadows. (Lack of shadows makes any solid object appear unnaturally "flat".)
• Motion parallax. Camera motion, tracking and panning help us tell which objects are near and which are far.
• Selective focus.

Selective focus may deserve comment. In the real world this is only an effective depth cue for nearby objects, a few inches away from your eyes. Hold a finger a few inches from your eye and focus on it. The more distant background objects will appear unsharp. In 2D photography limited focus depth is sometimes used in other situations to direct attention to a sharply focused object of interest. Stereographers generally avoid this and try to keep everything in the scene in sharp focus. The reason for this is that when viewing a stereo still picture, the viewer likes to move his eyes around the picture, looking at objects both near and far, expecting to be able to fuse them in sharply defined 3D. Blurry 3D isn't effecive.

Convergence and focus conflict.

In normal vision your two eyes converge on the object of interest, and the eye lenses adjust focus to that object. These convergence and focus functions are learned from birth, and are interlocked. It isn't easy for us to converge at one distance while focusing at a different distance. When one free-views a stereo pair, by either the parallel or cross eyed method, we must do this unnatural act. Most people can learn one or the other method, but some find it difficult, and tiring. Even when using an old-fashionend Bates-Holmes stereoscope (or other two-eyed optical viewing system) our eyes must to some extent "unlock" the focus and convergence that has become an unconscious function of our eye-brain system. The same is true for stereoscopic motion pictures.

The ortho factor.

When viewing any photograph, painting or drawing, whether in 2D or 3D, the greatest realism is obtained when each familiar object in the picture subtends the same angle to the eyes as it did in the real scene. Except for huge murals, dioramas, or IMAX movies, this is seldom achieved. If the angular sizes of objects in the picture are equal to those in the natural scene, we say the ortho factor is 1—the ideal situation. ("Ortho" is short for "orthogonal".) A postcard-sized picture is only a miniature representation of the scene. 8x10 inch or larger photographic prints aren't much better. Widescreen movies can come close to reproducing size relations accurately, if viewed from a distance of about one screen width from the screen. Most people don't sit that close.

At this point you might like an illustrated introduction to stereo geometry at Life and Depth, and introduction to Stereo Photography.

Stereoscopic depth and stereoscopic distance

The apparent depth or solidity of an object in a 3D picture is a function of the angular difference (disparity) between the left and right images of each object in the scene, as the stereo pair is presented to the eye. This disparity also governs the apparent distance of objects. But by changing the separation of the left and right eye pictures of a stereo pair we can unlock the relation between these two and make an object appear nearer, even though its stereo depth is "wrong" for that distance. This can cause apparent distortion of the apparent depth of objects, which is disturbing if the objects are familiar, such as people. Free-viewing methods, especially cross-eyed viewing, necessarily have this problem. Surprisingly, it does not bother people as much as you might expect, for our eye-brain system can adapt, and ignore, such conflicts (especially for familar objects). But it certainly reminds us that we are not seeing the real scene, but only a simulation of the scene.

Can we really recreate reality?

• Take two photos from points 63.5mm apart, with wideangle lenses, to duplicate the visual fields seen by each eye.
• Present these pictures separately and simultaneously to the eyes, enlarged for an ortho factor of 1.
• Present these images separated left/right so that all objects in the scene have exactly the same stereo parallax they had in the original scene.
• Use a large picture format that places the picture boundaries (the stereo "window") far enough removed from the center of interest so that the viewer is not even aware of them.

Now consider the present state of motion pictures. We have wide screens, and the largest IMAX screens would be our first choice for stereoscopic display. We have rather good ways to separate the left and right eye pictures, using circular polarization, but "ghost images" are still a nuisance for bright/dark boundaries of objects nearer or farther than the screen distance. Moviemakers try to avoid such situations when making the movie, and with computer digital techniques (ghost-busting software) they can sometimes be minimized even after the movie has been photographed.

What problems remain? Ask a seasoned stereographic movie maker what problems he must continually be aware of.

The stereo window. Most 3D viewing methods present the picture with a frame, and the frame has a firm location in space. Consider stereos viewed on your laptop computer, within a frame, say 20 inches wide and 20 inches from your eyes. Now imagine looking at everything through an empty picture frame of that size and location. You'd see only a limited segment of reality. And your eyes would have their focus locked at 20 inches, while stereo parallax could locate images anywhere from 20 inches to 20 feet. This can be tiring for many people if done for a prolonged time.

Fig. 4. The space of 3D cinema. A cone with its vertex at your eyes, extending to 50 feet away. The screen, S, is 20 feet or more away. P is positive parallax space. N is negative parallax space.

So what if the frame were much larger, and farther away, say the screen of a movie theater? Then the eyes would focus at the screen, perhaps 20 feet (or more) away, on images having stereo parallax from 20 feet to about 50 feet. Anything beyond 50 feet has very small parallax and appears to be at "infinity". This is true even in the "real world" we see. This is the limited slice of space that stereographers have to work with.

But screens at most theaters are still not large enough, so the audience is continually aware of the presence of the location of the screen edges in space. Objects of interest in the scene should not seem to intersect the left and right edges of the frame, especially when those objects are to appear in "negative space" (appear nearer than the screen). Stereographers call this a "frame violation". Now consider that the screen (and the picture frame) are some 20 feet away from the viewer in the center of the theater. In everday life the stereoscopic function of our eyes is most effective at 50 feet or less. Consider an intimate scene in someone's living room, or other smaller space. To reproduce it effectively in a theater would require almost everything in the scene to appear nearer than the movie screen.

But that would present all sorts of stereoscopic problems. Ghost images, frame violations, and skew distortion (for people sitting off to the sides of the theater). This is usually avoided in 3D movies. So the moviemakers generally contrive to adjust the separation of the two images so that objects of interest appear at or slightly beyond the distance of the screen. But in the theater that living room scene may appear to be over 20 feet away, and everyone appears too small.

You might wonder why moviemakers don't move the apparent window closer than the screen itself, allowing action to utilize the "negative space" between the screen and your eyes. That has been tried, but the frame is a boundary between light and dark, so ghost images of it become a serious distraction.

To retain a sensation of realistic depth in such a situation is difficult. L/R images of objects very distant should not be separated by more than 2.5 inches on the screen, or viewer's eyes will be forced to diverge, which is not something our eyes easily adapt to. Headaches and eyestrain can result. So, you see the moviemakers have only a small range of image separations with which to make the stereo effect happen.

This is why much of a 3D movie seems "flat" to many people. There just isn't much depth except in a few brief situations where depth is used for dramatic or shock effect. But if that is over-used, it can annoy audiences.

I've even heard some complain, after seeing Avatar 3D that after a half hour or so, they forgot they were watching a 3D movie. Then upon leaving the theater the real world seemed "flat". This just reminds us of the importance of human psychology and human expectations, and the fact that the sensation of 3D reality is more than just 3D parallax. We are seldom consciously aware of stereo depth in everyday life, but film makers feel an obligation to make us constantly aware of it in a movie. And movie patrons expect it, or they judge the 3D movie "not worth the extra cost."

Where shall we sit? The stereo illusion very much depends where you sit in the theater. Suppose the movie has a scene with an object intended to appear halfway between you and the screen. That distance is much less for someone in the front seats than it is for someone in the very back seats. Sitting farther back "stretches" the spatial z-axis, sitting in front shortens it—for all objects in the scene. Sitting off to the side introduces an interesting skew distortion of solid objects, so a cube is no longer a cube, but more of a parallelopiped. Reality is warped. At several theaters I attend, the ticket sellers tell each patron of a 3D movie to "Sit near the center of the theater for the best picture."

Stereo convergence. The human eyes are never at rest. They move to scan the visual scene. And the head moves, too. Can the photographic stereo presentation allow for this in a natural way? Of course the motion picture camera can move, panning and even zooming.

In real life we concentrate our gaze on something that interests us, and usually adjust our head position to match that. The visual field is much larger than the field of interest. Objects at the edge of the visual field are seen in much less detail and resolution. In fact, stereo rendering is poor at the edges of the visual field. But in a theater, not all patrons will look at the same object of interest. The projected image must allow good stereo rendering over the entire field. Of course the photographer may compose the scene to "force" most people's attention to a partucular place.

Some photograhers employ lens axis convergence to place objects of interest at the distance of the screen or greater. This invariably distorts objects at the edge of the screen. It causes vertical displacement of the two images. For this reason many photographers feel that axis convergence of the stereo lenses should be avoided. There's no doubt that convergence annoys some viewers. The alternative strategy is to decrease stereo lens separation, and this creates its own problems, making objects appear subjectively too small and distant on the screen.

Contrast and gamma. Throughout the history of cinema moviemakers have sought to achieve the effect of natural tonal balance in scenes. Never have they achieved a full range of image brightness in the theater, matching that in nature; not even matching the range our eyes can comfortably perceive. If movie makers expose for the highlights, the shadow detail suffers, and vice versa. If both highlights and shadows have detail, the picture seems to have compressed brightness range. The situation is no better with digital photography. And with stereo movies, the very low light efficiency of the process (due to the necessary filters to keep L and R images separate) requires very high intensity projection lamps and high gain reflective screens. Still critics complain the 3D movies appear "dark" and dull compared to the same movie seen in a 2D theater. In fact, digital 3D picture content is computer-processed to reduce this problem by reducing the gamma (the difference between light and dark) and reducing contrast. This is a step backward in picture fidelity, a sacrifice to make the 3D work.

Test it yourself. During a 3D movie, remove your 3D glasses. You may find the screen brightness dazzling, if your theater is using the highest intensity projection lamps. If you are not dazzled, then your theater is cutting corners (and its electricity bill) by using inadequate lamps.

Spurious reflections. I've been in some theaters that had annoying ghost images displaced from bright areas of the picture. I concluded that these were from the glass windows of the projection booth. In the days of film, projectors were somewhat noisy, and glass projection booth windows was a way to block this sound from the back rows of the theater. Nowadays digital projectors are quieter, and these windows are not needed. When I complained about this to one theater manager, she said it would be difficult to remove the glass. I suggested that a hammer might do the job.

Ghost stereo images. Polarizing glasses aren't perfect. Some of the right image reaches the left eye and some of the left image reaches the right eye. This is usually not noticeable except for very bright objects against a dark background. Careful set lighting can avoid this. Digital motion pictures can be computer processed with "ghost-busting" software to identify and minimize such problem areas of the image.

Will 3D movies please everyone?

Some people have two fully functioning eyes, but their brains don't process the images to 3D, even in everyday life. They could cover one eye and not feel disadvantaged except when doing certain tasks requiring eye-hand coordination.

And there are those (I have no idea how many) who experience eyestrain or even headaches in 3D movies. Perhaps it is a result of the unusual effort of unlocking eye focus and convergence.

Certain film critics, Roger Ebert being one, are negative toward 3D movies, and ask "Have you ever seen a 2D movie that you thought would be better in 3D?" My answer is "Yes, most of them." But obviously not everyone shares my view. I'll admit that I've seen some 3D movies that would have been just as good (or bad) in 2D.

The moviegoing public is largely technologically illiterate about 3D. I've encountered some people who thought you could save the 3D glasses and then use them on a 2D film to see it in 3D. (That would just darken the picture.) Some think these glasses can be used as polarizing sunglasses. (They are not suitable for that.)

Unnecessary problems with 3D.

But now that many theaters are installing digital projectors and metallic screens, 3D is shown using only one projector. Image alignment is done at the movie studio, and can't be compromised at the theater. So that should remedy the presentation problems of the past.

The technology of today's 3D exhibition systems are capable of very good results. Unfortunately, sometimes the results are compromised by carelessness of theater personnel. There have been instances where audiences were forced to suffer through presentations that reversed the left and right eye pictures because someone hadn't flipped the right switches on the digital 3D projector. In one of these cases someone in the audience was clever enough to put on his glasses upside down, and others in the audience did the same, which corrected the problem. In cases such as this there is often no one in the theater knowledgeable enough (or with authority) to fix the problem. A technician must be brought in from outside, which could take a day or so. I have also experienced 3D movies in which the picture was stretched horizontally, making characters appear too fat and circles become ovals. This took over a month for the theater to correct. Moviegoers may not realize that theaters today may not even have a projectionist on duty. The equipment and movie content is set up by the theater manager for maybe 10 screens at the start of each week, and then left to run mostly unattended, except for occasional replacement of a burned-out lamp. My advice in such cases of poor exhibition is to complain to the manager (if one can be found), not to one of the clueless teenagers who sell the tickets and junk food.

Recently, I saw a 3D movie in a theater that had previously had good digital 3D presentations. The movie seemed somewhat "unsharp" throughout. Looking carefully as the end titles scrolled I realized why. The picture was sharply focused only at the top edge, and got progressively "fuzzy" below and was worst at the very bottom of the screen. This is a problem that a competent projectionist could easily remedy, or avoid in the first place. But this theater had no projectionist, and when I complained, the manager said she would call in a technician the following day. I am convinced that many people's negative feelings about 3D movies result from exhibitor incompetence, not from the technology of 3D, and not the fault of the film producers. From previous experiences, I'm not impressed with the competence of these "technicians".

There are still some movie producers who have poor judgment. Even the legendary director James Cameron has been guilty of using 3D camera axis convergence, though he seems to have mended his ways in his technically excellent movie Avatar 3D. But even there, in some early live action scenes, he had shots in which distant objects had their L and R images several feet apart on the theater screen, which is certain to induce eyestrain and headaches.

3D movies with 1D plots. If the current wave of 3D movies fizzles, it may be the fault of content, not technology. We've had two years of 3D movies now, mostly computer-animated ones with plots aimed at children or teenagers. How many fully live-action stereoscopic movies have there been with adult plots? There seem to be very few live action 3D movies planned for the coming year, and when we do get some, they will probably be action films with comic-book superheros or sci-fi flicks. I'm not optimistic for the artistic future of 3D if this trend continues.

Stereo pictures and stereo sound; a comparison.

The bottom line.

A few years back we saw a fad of "computer-colorizing" old black/white movies. That seems to have run its course. Now we are seeing 2D to 3D movie "conversions" done largely with computer software. Most people can't tell the difference between a converted film and one conceived and produced in 3D. Let's hope this fad soon dies as critics and moviegoers learn to appreciate the subtle differences. I'm not denying that a colorized and 3D converted version of Casablanca or Citizen Kane might be interesting to watch. I would pay to see that. But let's not forget, or fail to appreciate, the artistry of those who made great art working within the technical limitations of the media available.

—Donald E. Simanek.

Return to top of this page.
Return to the 3D and Illusions page.
Return to front page and main menu.