Creating Good Stereo-Vision Images

Introduction

StereoGraphics has a great interest in helping software developers to produce the best-looking stereoscopic images possible. Not only should such images have a good depth effect, but they must also be easy on the eyes. Both of these concerns can be satisfied, but it takes some experience to recognize a good stereoscopic image, or as important, a poor one.

When looking at a planar image on a display screen, it doesn't matter whether you're looking at it with two eyes or one eye: Shut one eye and there's no difference in the effect. But shut one eye when wearing CrystalEyes® or SimulEyes VR eyewear when looking at properly prepared software, and there is a substantial loss of depth effect. The only difference between our display and an ordinary or planar display is these two slightly different perspective views. With an electronic stereoscopic display we are presenting both eyes with the same slightly different perspectives they have when seeing the real world.

The key element in producing the stereoscopic depth effect is parallax. Parallax is the horizontal distance between corresponding left and right image points. The stereoscopic image is composed of two images generated from two related perspective viewpoints, and the viewpoints are responsible for the parallax content of a view. We'll return to a discussion of parallax, but first some remarks on how to heighten the stereoscopic effect.

Good Looking Effects

Stereoscopic images tend to look deeper when extra-stereoscopic, or monocular, depth cues are stressed. The major cue that adds to stereoscopic depth is perspective, or the relative juxtaposition of foreground and background elements of a view. So, when creating images, if your camera model uses a wide-angle lens you will stress the perspective cue, because the size of objects close to the lens will be exaggerated compared to objects which are farther from the lens.

Another effect you can use to accentuate depth is called "motion parallax," and once again this has to do with perspective changes or relative juxtaposition of foreground and background. These perspective changes occur when there are certain kinds of camera moves. For example, forward motion of the camera will create juxtaposition of near and far objects. (This is different from a zoom. In a zoom, you're changing the focal length of the lens without changing the position of the camera or the lens, and there will be no perspective change.)

If the camera lens is looking perpendicular, or at some angle to the direction of the camera travel, there will be relative juxtaposition of foreground and background elements, and this will heighten the stereoscopic effect. (This should be distinguished from a pan or tilt. In a pan or tilt, the camera and lens rotate and view the three-dimensional space without any relative juxtaposition of foreground and background.)

There are other important depth cues that can be used to heighten the stereoscopic effect. One is relative size. Objects closer to the lens will be larger. As mentioned earlier, this effect can be exaggerated when using a wide-angle camera model. Color can also be an effective depth cue. Brighter, lighter colors -- reds and yellows -- tend to look as if they're closer than darker blues and greens which tend to recede. In addition, shadows and shading play an important part in modeling the scene.

Once the principles are explained, there is really no way to quantify most of this. It takes the eye of an artist with good aesthetic judgment. The software developer should be aware of these factors and their importance. Parallax, on the other hand may be quantified, and we will discuss it below.

Parallax

When looking at stereoscopic images on a computer screen, large parallax values may cause discomfort to some viewers. There are two kinds of parallax we're talking about: Uncrossed-parallax and crossed-parallax.

Uncrossed-parallax has the left corresponding point to the left of the right corresponding point. Uncrossed-parallax will produce image points which appear to be within the monitor itself. On the other hand, crossed-parallax has the left image point to the right of a right image point, and such image points will appear to be coming off the screen. Zero-parallax points, in which the left and right corresponding points or objects entirely overlaps, will appear to be at the plane of the monitor screen.

In terms of allowable parallax values for uncrossed-parallax, for the typical computer screen in which people are going to be sitting two or three feet away, try not to exceed about half of an inch of parallax. Since parallax is a function of magnification, the bigger the monitor the larger the value of parallax. So lets establish, for the sake of argument, a standard monitor size of 17 inches with half an inch of parallax. Smaller or larger monitors with the same software will have smaller or larger values of parallax, but will probably be within the acceptable range.

Crossed-parallax will make objects appear to be coming off the plane of the screen. You probably don't want to exceed values of more than about an inch or so for objects that are going to remain in the field of view for a long period of time. For objects shooting off the screen rapidly, you can have large values of parallax. If the object is going to be leaving the screen surround (the four sides of the screen), it can have very large values of parallax as it leaves the screen which makes for a dramatic effect.

Avoid crossed-parallax objects which are cut off by the screen surround, especially the vertical left and right edges. This tends to look weird to most people because there is a conflict of cues. The stereoscopic cue tells you the object is in front of the plane of the screen, but the fact that it is cut off by the surround tells you it is behind the plane of the screen. There are some ifs and buts to this. In many cases, objects can be hung or spring up from the horizontal edges of the surround. For example, you can hang a chandelier from the horizontal edge and no one will mind; it will probably look just fine. Or, similarly, you could probably stick a pole up from the lower horizontal edge.

One thing that's fun about stereoscopic displays has nothing to do with the depth effect. Rather it has to do with glitter, sparkle, and luster. These effects are related to the fact that certain objects refract or reflect light differently to each eye. For example, looking into somebody's eyes, or at a diamond, there may be highlights that can be seen in one eye but not in the other. This is a fun effect that can be added to a stereoscopic image. It also is an effect that can be used for flames and fire to make them look alive. For example, if you use a bright orange for the fire in one eye and a red in the other, you will create the effect of scintillation.

Conclusion

Producing stereoscopic images for an electronic display screen is a challenge. But the effects can be fun and the images can be beautiful. It's an art, requiring judgment and skill.