three-dimensional photography

In any scene there are clues to its three-dimensionality, such as variation in apparent size between near and distant objects, overlapping of distant by near objects, and aerial perspective, but the most powerful clue is parallax—the positional changes in the scene as the viewpoint is changed. The left and right eyes are on average some 65 mm (2 1/2 in) apart, and receive slightly different images of a scene as a result of the differing viewpoints. This is interpreted by the brain as ‘depth’. In real-life situations the sensation is not strong, as it is swamped by the much greater dynamic parallax resulting from physical movement. A conventional photograph is static, and has no parallax, as it has a fixed viewpoint. But if two photographs are taken from positions 65 mm apart, and the resulting prints are examined via an optical device that presents each image separately to the two eyes, normal static parallax is introduced, and the image appears to have depth. The principle is known as stereoscopy and the photographs are called a stereoscopic (or simply ‘stereo’) pair. Accommodation (visually focusing), a further clue to depth, is not involved, and ocular convergence for different distances (another clue) does not usually match that which would have been involved in viewing the original scene. Nevertheless, to people with normal binocular vision the sensation of three-dimensionality in the image is powerful.

Hand-drawn stereoscopic pairs appear to have existed before the invention of photography. The first stereoscopic viewer for photographs was designed by Charles Wheatstone (1802-75), and subsequently improved by David Brewster. The stereoscope was a popular toy in late Victorian times, and many stereoscopic image pairs still exist.

Stereoscopic camera formats

It is possible to produce stereo pairs with a conventional camera by using a platform on a tripod, with a guide to slide the camera along between exposures, but this confines the technique to static subjects. Stereoscopic cameras with twin lenses have been around since the early days of photography, and several versions are available. At least one specialist in Britain produces Siamese-twin modifications of professional SLR and digital cameras, and several camera manufacturers offer dual-prism conversion sets for the production of stereo pairs in portrait format with conventional 35 mm cameras. The Nimslo camera of the 1980s employed four lenses to produce a somewhat crude lenticular stereogram, and was initially popular, but its optical limitations, and logistic problems related to processing, eventually led to its falling out of favour.

Hyper- and hypostereoscopy

When the subject matter is distant it is often beneficial to lengthen the stereo base, to enhance the impression of depth slightly. This is called hyperstereoscopy. An extreme example of this occurs in aerial survey photography, which utilizes a series of overlapping vertical photographs made during straight and level flight, with a spatial separation of up to 500 m (1, 640 ft). When pairs of images are viewed stereoscopically through appropriate optics, objects on the ground appear small, but heights appear exaggerated because of the excessive parallax angle. While aesthetically unpleasing, this is a positive advantage in map- making as it allows greater precision in contour plotting.

In contrast, some stereo adaptors for hand cameras give a base less than the average interocular distance, so that the images of objects at normal distances appear somewhat flattened (sometimes called ‘cardboarding’). However, a short base (hypostereoscopy) is mandatory in photomacrography, as the parallax angle should not exceed about 7 degrees for a realistic 3-D effect. In microscopy the camera is usually fixed, so instead the specimen is rotated between exposures through 5-7 degrees. The result is an apparently large image some distance away, but this is usually acceptable in this type of image.

The Brewster stereoscope, one of the oldest viewing devices, uses prints of a suitable size set up side by side, with a pair of simple convex lenses to put the two images at infinity. For larger prints there are more sophisticated designs using pairs of mirrors or rhomboidal prisms to cope with the increased distance between the print centres. With practice it is possible to view a small stereo pair without optical aid, by training the eyes to focus without convergence. A more modern technique dissects and interlaces the images, allowing the viewing of much larger images in a similar way. Stereo images that do not require optical aids are called autostereograms.

A type of stereo image needing less personal effort in the viewing is the anaglyph. The left and right images are printed superposed in respectively blue (or cyan) and red, and are viewed through spectacles coloured respectively red and cyan. It is possible to use this technique for images in natural colour, the left eye receiving the red image and the right the green and blue, but not all viewers find the result satisfactory. A better method involves two full-colour images polarized at plus and minus 45 degrees and viewed through similarly crossed polarizing spectacles. Both systems are suitable for projected images, and have been used for 3-D movies, but the latter technique demands a special metallized screen.

Full parallax systems

The ultimate in three-dimensional imaging is holography. The lenticular stereogram is a full-parallax photographic imaging system. This initially employs a moving camera (or a multi-lens camera) to record a whole series of images; these are dissected and printed as interlaced strips, displayed behind a lenticular screen—a grid of retroreflecting cylindrical lenses that operate like a catseye reflector. The result is that each eye sees only the view recorded from that particular viewpoint. Dynamic parallax thus augments the stereoscopic imagery, and depth is perceptible even with one eye closed—albeit over only a fairly small angle. Unlike holograms, photographic stereograms do not need special illumination; they can be made in full colour and any size, and several people can view the image at the same time. But they still lack some of the requirements for the full 3-D experience, for example, ocular accommodation.

Further developments

There are a number of current developments involving computer programs that change the perspective of an image as the viewer moves around, as well as others that exploit visual illusions. There are also systems that produce genuine three-dimensional images (‘integral imaging’) using complicated arrays of microlenses. But for the present, lenticular stereograms present the most convincing form of 3-D photographic image.

— Graham Saxby

Bibliography

• Okoshi, T., Three-Dimensional Imaging Techniques (1976).
• ‘Three-Dimensional Image Capture’, SPIE Proceedings, 3023 (1997).
• Ray, S., Scientific Photography and Applied Imaging (1999).
• Reynaud, F., Tambrun, C., and Timby, K. (eds.), Paris in 3D: From Stereoscopy to Virtual Reality (2000).
• Saxby, G., The Science of Imaging: An Introduction (2002)