Foundations of the Stereoscopic Cinema
by Lenny Lipton
Pub. Van Nostrand Reinhold, 1982, 319pp.. £18.65.

REVIEWED by CHARLES W. SMITH

The BKSTS Journal February 1983 p 72
Until now there has been no book of any kind available on the technique of 3-D film production; the well-known "Theory of Stereoscopic Transmission" by Raymond and Nigel Spottiswoode (pub. University of California, 1953) which dealt chiefly with theoretical considerations has long been out of print. Lenny Lipton's new book therefore comes at a timely moment to fill the void.

This turns out to be rather a strange publication for what at first glance has the appearance of a technical manual. It is like a television programme, frequently broken into by commercials praising the excellence of the author's work and the qualities of his present-day business associates. The pronoun I is in constant use, which is not customary in technical writing. Lipton is the sort of writer who is able without the least embarrassment to write of one of his own amateur 8mm films as "one of the finest stereoscopic films ever made" (he does not quote any independent testimony).

The reader soon discovers that Lipton's knowledge of the subject is very limited, for anyone undertaking a full-length book. His background is that of a photographic journalist, and at the time of writing his practical experience of 3-D was limited to his own 8mm test films, projected on a 4ft screen. (He gives illustrations of his camera and projection set-up). He has never seen a Russian 3-D film, or a British one for that matter.

Of the four integral stereoscopic movie cameras built so far, Lipton has illustrations of the two American ones, but he is unaware of the existence of the two European ones. He has however made a thorough search of the files of the U.S. Patent Office and also of the SMPTE Journal in what he calls "exhaustive research" (though this didn't extend as far as the files of the BKSTS Journal).

In order to justify the publication of a new 'system' for 3-D filming, Lipton has found it necessary to attempt to discredit the work of the Spottiswoodes, claiming that it is erroneous, or superseded by later discoveries. This he does by a technique, not of quoting from the Spottiswoode's book, but by putting in his own words what he claims to be their opinion, and then attacking it. This is the novel section of the book, and it certainly merits close attention from those who now and in the future will be working in the 3-D medium.

The Spottiswoode Analysis
Apart from thinking it difficult for film-makers to understand, Lipton levels three charges at the Spottiswoode analysis: first, that they make the assumption that the eyes function like a rangefinder, by which Lipton means (p. 112) that distance is assessed by the muscular effort of the convergence muscles; secondly, that they disallow divergent screen images which require the eyes to squint outwards; and thirdly, that their analysis is based on Euclidean geometry, which "on a cosmic scale" has now been shown to be incorrect. These charges are returned to at frequent intervals in the book, so that "Spottiswoode" has far the most numerous entries in Lipton's index.

To check the accuracy of these charges, I have been looking again into the Spottiswoodes' "Stereoscopic Transmission", to refer to their own words. To take the first charge: If we look at their chapter "On the Perception and Transmission of Depth" we find (p. 13): "The disparate retinal images which enable the two eyes to fix a distant object, rangefinder fashion, rely upon the relative displacements or parallaxes of outlines and inlines in the scene which are either vertical or have a marked vertical component". It is clear that they attribute the sensing of depth to the extent to which corresponding image points in the two eyes fall on either corresponding light-sensitive cells in the two retinas, or on neighbouring but non-corresponding cells.

Later (p.56) they put it more concisely: "Retinal disparity produces the binocular impression of depth". This I believe is still the accepted explanation of binocular depth perception. To write as Lipton does (p. 141) that it is an implicit assumption of the Spottiswoode transmission theory that convergence of the eyes is a depth cue, and that this has been shown to be false, is therefore a distortion of the truth.

Whatever the mechanisms of depth perception, there is certainly no doubt that binocular parallax gives precise and accurate placement of projected screen images. This is easily verified with stereo images from a pair of slide projectors, when the position of a forward image can be readily matched against a pointer held by a colleague. As the convergence of the projectors is altered, the stereo image position moves forward and backward, in strict accordance with geometrical theory. This firm image placement is well-known to anyone viewing a 3-D film who has identified a forward image as being precisely over the heads of the audience, perhaps precisely two rows in front; and similarly to anyone who has looked through anaglyph glasses at a 3-D comic, or assessed the 1/4" depth of a Nimslo snapshot extending above and below the surface.

Psychological Factors
It is true that the mental interpretation of an image does not necessarily follow precisely its geometrical position because of psychological factors, and in particular that forward images which intersect the sides of the screen or (to a lesser extent) the top or bottom are held back by the conflict between the binocular information and the logical information that anything disappearing at the edge of screen, the stereo window, must be beyond that distance, or it would still be visible.

The Spottiswoodes were of course fully aware of these considerations, and the last section of their book is entitled "The Human Factor in Stereoscopic Transmission". In introducing this subject they write (p. 143) "The human mind, however, does not interpret spatial relationships solely on the stereoscopic hypothesis. It is able to fit together information received from many different sources, shifting, comparing, rejecting, and finally transmitting to the brain a statement which is much more sophisticated than the first crude message which it received from external sense data."

Lipton's accusation that the Spottiswoode analysis is based on Euclidean geometry is certainly correct; however he does not suggest any satisfactory alternative, and his own computations for the depth tables later in his book are equally Euclidean; so the complaint seems rather pointless. Although 'on a cosmic scale' light does not travel in straight lines, it certainly does so for all practical purposes within the range of stereoscopic vision, up to two or three hundred yards. If light didn't travel in straight lines, the projection of motion picture film would be impossible.

It may be that Lipton is here confusing the placement of the stereoscopic image with its mental perception. Rays from the image and from the real world have both to pass through the process of perception. The requirement of the stereoscopic image is that its rays should simulate as closely as possible rays which could be expected to arrive from a similar object in the real world.

Divergent Images
Lipton's third charge, that the Spottiswoode system disallows divergent images, deserves closer consideration, since the extent to which these are to be permitted is a basic decision which has to be taken by any technician put in charge of the stereoscopic image control of a 3-D film. Divergent images require the spectator's two eyes to squint outwards. This never occurs in the real world, so the eyes are not accustomed to it, and this is known to be a cause of headache. However there is a certain tolerance before the effort of fusing divergent image points becomes painful, which doubtless varies from one spectator to another, possibly to some extent related to the frequency with which they have viewed stereoscopic films.

Lipton quotes with approval the estimate by Valyus (Stereoscopy, 1966) that the limiting amount of divergence, at which stereoscopic fusion becomes impossible and the scene breaks down into two separate images, is reached when the difference between the angle of divergence and the eyes' convergence on the screen plane (the plane of focus) exceeds 1.6°. Valyus does not describe the experimental work on which this figure of 1.6° is based.

In order not to exceed this outside limit, Lipton decides to limit divergence of the eyes on background points to 1° only. This is in fact a large angle in stereoscopic terms, since the angular difference between convergence on the screen and the parallel condition amounts to 0.4° for an observer sitting 30ft from the screen. If 1° of divergence is to be permitted, then for our 30ft spectator, over two-thirds of all rear-screen images will be divergent.

The extent to which divergent images, even when they can be fused, give rise to discomfort after extended periods of viewing has not yet been tested. All stereographers seek to limit divergence to unimportant background points which the brain will not wish to scrutinise closely. Lipton himself writes: "Divergent homologous points can prove to be troublesome when shooting close-ups that have a distant background". Also: "To be compatible with the creation of a stereoscopic effect, image points should have the lowest possible screen parallax". And also (p.191): "If the composite requires the viewer to observe the background in preference to the foreground, then divergence ought to be avoided". This seems remarkably close to the Spottiswoode position, which we will quote shortly.

But the trouble is, as Lipton says a little plaintively, that "it is far simpler to do photography with large values of K (screen Parallax) than small values". Advocates of divergence are thus open to the charge of putting the convenience of the cameraman before the comfort of the spectator. Lipton says (p. 103) "Most people's eyes can accept a small amount of divergence without strain"; which is certainly a guarded way of advocating divergence. It is to be noted that the advocates of divergence on background points do not claim it to be advantageous, but merely that within certain limits it is tolerable.

Reduced Interaxial
In practice, the only means of avoiding divergence on background points behind a close-up is by suitable reduction of the interaxial separation. Extreme close-ups require extremely small interaxials, so that a rig with zero-separation facility becomes desirable. And so we find that stereographers who do not have access to small-separation cameras argue in favour of divergence as desirable; stereographers with small-separation rigs on the other hand usually argue that divergent images should be avoided. Lipton comes into the former category, since he tells us his 8mm rig had a minimum lens separation of 66 mm, about the human eye separation, and his later work has been done with the fixed-separation Stereovision lenses.

To turn now to the actual Spottiswoodes' recommendation (which Lipton does not quote): they wrote (p.33) "It is found that divergence is likely to cause eyestrain, and therefore screen parallaxes in excess of the eye separation should be avoided". But they also went on to say, in listing future development requirements, that "Much experimental work must be carried out to determine limiting values of divergence at different viewing distances which are acceptable without eyestrain".

Raymond Spottiswoode had responsibility for preparing and presenting the programme of 3-D films at the 1951 Festival of Britain exhibition. He therefore had the opportunity (unlike Lipton) of putting his films before a paying audience,, on a full-size screen, and taking sample polls of audience reaction. He gave high priority to viewing comfort for the audience. Many people think it would have been in the better interests of the industry if later producers had concentrated more on viewing comfort and less on 3-D sensationalism.

Depth-Range Formulae
We must turn now to Lipton's own proposals. He too derives depth-range formulae, but under two different headings: first, for background images free from divergence, and second, with greater depth range, allowing divergence up to 1°. The first figures are recommended for use in scenes where the background will be subject to scrutiny; the second for scenes where the background is unimportant, a mere backing to significant foreground detail. So the difference from the Spottiswoode position is again only one of emphasis.

As a conscientious reviewer I tried hard to follow Lipton's derivation of his formulae, but found it hard going. Unfortunately he has not numbered his equations, so it is sometimes difficult to know what previous result he is referring back to. Also, equations are sometimes carried forward incorrectly; as for instance the upper equation on p.202, misquoted from the previous page.

A new mathematical nomenclature has been adopted for the various optical factors of image and object distance, lens focal length, interaxial separation, and so on. In some cases Lipton has followed Spottiswoode, in others preferred to use a different letter. Very often he gets confused with his own
nomenclature so that results become incorrect or incomprehensible. As an instance, in the top four equations on p. 199 the terms Dme and Dmd have got reversed, giving the absurd result of a scene where the depth available with divergence is less than the depth available without divergence.

Errors
Many examples of such errors could be given. Without labouring the point too much, here are a few. P. 195, 2nd equation, Dhd should read Dhe. P. 116, last 2 equations, p should read Dm. Diagram 3.10, angle LB'R should be a1, not a. P.202, 3rd line, Dh should read DHE. P. 234, near bottom, Dme and Dmd are quoted wrong way round. P. 244, penultimate paragraph, 'D2 should read D2. Table 8.3 is incomprehensible because it introduces a new factor N which (so far as I have been able to find) is nowhere explained. On pages 116 & 117 alone I have noted six mathematical misprints. Quite a lot of detective work is needed to try to make sense of the argument.

Lipton's algebra also seems shaky. On p. 216, the derivation of the third equation from the second is incorrect. This doesn't seem an error that can be blamed on the printers, since a similar mistake is found in the first two equations on p. 114.

Twelve pages of depth-range tables are given, for 8mm, 16mm, 35mm and 70mm films, with 3 focal lengths for each film size, and a choice of five interaxial separations between 25mm and 105mm. It will be seen that these tables are not of great practical use, since the range of five widely-spaced interaxial separations would not be adequate for anyone with a variable-separation camera. Since Lipton tells us the only variable-separation camera he has used is his 8mm rig, with no separation available smaller than 66mm, the tables are anyway only suppositional. Lipton in fact advises his readers to compute their own tables; the examples he gives may be regarded as a guide to one possible form of presentation. He declines to follow Spottiswoode in the use of reciprocal distance units, although this greatly abbreviates depth-range tables since the depth is then the same value for all object distances.

The tables do not give any values for the required convergence in angular (or other) measure for the given lens separation and object distance. Lipton seems to assume that convergence will be set by the
primitive method of lining up an object at the required distance successively on the ground-glass crosslines of the two cameras.

Stereoscopic depth tables are only valid for a single size of screen. The screen sizes for which Lipton's tables are worked out can be deduced to be 10'3" for 16mm film, 25' for 35mm film and 58' for 70mm film, although he omits to mention this.

Shape Distortion
One of the striking originalities of "The Theory of Stereoscopic Transmission" was the analysis of the shape reproduction to be expected of images viewed in a stereoscopic system. The authors pointed out that the depth magnification of an image, as compared to the object, would in general not be the same as the width (and height) magnification; they devised the concept of the Shape Ratio, which is Depth Magnification divided by Width Magnification, to assess the shape of the image; and they distinguished between scenes in which the shape reproduction would be constant throughout the scene, and those in which it would be nonlinear, differing in the foreground from the background.

The concept of the Shape Ratio has been found to answer well to the practical requirements of 3-D filming. It serves as a valuable guide in such tasks as the shooting of pack-shots for 3-D slide shows, where clients can be highly critical of any distorted representation of their products. In 3-D TV commercials this can be expected to be an important factor.

Lipton doesn't follow the Spottiswoode formulations on image shape; but neither does he offer any alternative advice. He does give a definition of Object Magnification, using the term to cover Width Magnification only; but he at once falls into the trap of confusing the stereo image size with the size of the 2-D screen image — the two are only the same when the image is in the plane of the screen.

At this stage, Lipton abandons his formulae altogether, falling back on vague generalisations — p.220: "The cues of stereopsis and perspective can be made to work together to produce pleasing images", etc.

Anyone wishing to make tests of the shape characteristics of stereoscopic images I would strongly advise to work with 35mm slides, not 8mm movies. Slides are cheaper and simpler, the definition is much better, a much bigger picture can be projected, the convergence can easily be altered by swivelling the projectors, and every scene can be held on the screen as long as desired, whilst the image is examined from different viewing positions.

Since his 8mm days, Lipton has graduated to become a professional 3-D Expert, and took technical charge of the 1981 feature production Rottweiler, shot with Stereovision lenses. Lipton speaks highly of his work on Rottweiler, describing it as "one of the best shot stereoscopic films ever produced". He clearly has high hopes for the success of this film, and that it will be acknowledged as revolutionising the quality of 3-D filming and thus prove the success of his 'system'.

Rottweiler has now had its initial showing, and subsequently seems to have been withdrawn from distribution. I asked recently in Hollywood what had happened to it; I was told it had been returned to the laboratories to have the stereoscopic errors corrected. This may perhaps be an unkind joke at Lipton's expense.

The 3-D Media
Although it has been necessary to draw attention to many errors, there is also much that is sound sense in Lipton's book. The historical sections arc fascinating, with many illustrations from early patents and amusing comments on their absurdities. One cannot doubt his intense belief in the future of the 3-D media. The book has many diagrams, clear but rather poor in quality by comparison with the current standard in photographic books; the author has not attempted to provide stereoscopic illustrations (as the Spottiswoodes did in an accompanying booklet of anaglyphs).

"Foundations of the Stereoscopic Cinema" fails to deal in any way with 3-D animation, or titling, or puppetry, or any of the special-effect processes as opposed to straightforward photography. It will certainly find many readers, since it is now the only book available on the subject, and there is just now a great upsurge of interest in 3-D image processes of all types. But readers would be well advised to verify the comments on stereoscopic image position and viewing comfort against their own observations, since it is known that psychological factors which influence the mental perception vary between individuals, and we are by no means dealing with an exact science.

Many readers will be surprised at the lack of credit given to other workers in 3-D; many will feel that the pages devoted to Lenny Lipton's early life could have been put to better use. The book ends on a triumphant note. In his last paragraph, Lipton announces that on November 20th 1981 he completed the invention of a splendid new system for high-quality 3-D television. Once again, we shall have to wait and see.

Foundations of the Stereoscopic Cinema: A Study in Depth
Lenny Lipton, 311 pp., illus., bibliography, index. ISBN 0-442-24724-9. Van Nostrand Reinhold Co., New York (1982) $21.95.
Reviewed by Stephen A. Benton, Polaroid Corporation, Research Lab., 750 Main St., Cambridge, MA 02139.

The current revival of interest in "3-D movies" is only the latest phase of a hundred-and-fifty-year history of attempts to bring the richness ol high quality "spatial imaging" (to embrace three-dimensional imaging in its widest sense) to bear on our everyday experience. Hopes for sustaining this revival lie in the more advanced photo-optical technology now more widely available and in the more sensitive and intelligent use of that technology, By drawing together a wide variety of historical, mathematical, and practical data, Lcnny Lipton works to provide a firm intellectual footing for independent film artists considering this enhancement of their medium. For the making of a satisfying 3-D film is much more than twice as complex as for a conventional film, and the many new decisions require substantial care. They should be based on technical understanding as well as experience, from the projectionist who must align and balance his equipment (at no extra pay!) to the cinematographer who must decide how best to separate and direct the lenses for the intended effect.

In the effort, Lipton provides two very different books, one much better than the other. A comprehensive bibliography (more than 350 entries) compiled by his associate Michael Starks is the basis for the first, a widely ranging, profusely illustrated (152 figures and tables), and engagingly annotated review of the history of spatial imaging's concepts and inventions. Lipton reveals a fond interest in the offbeat and quaint before focusing on the (relatively) practical technology of modern stereoscopic lilmmaking, and attempts to codify and compare six different approaches to stereoscopic filming calculations. This is a valuable reference, despite a few historical and technical flaws (and contentious opinions), but its attempt at overwhelming authoritativeness cannot overcome a lack of intelligibility and logic in supporting what follows.

It is a surprising fixation on mathematics that most erodes the "second book's" utility- surprising in view ol Lipton's own remarks that binocular depth perception is not only highly idiosyncratic but also varies markedly with the scene content.

Indeed, Lipton was so dismayed by the variability of his viewing audience's results that he discarded them in favor of his own observations! Even then he concludes that good 3-D is more an art than a science, which is only to say that our understanding of it is still too simplified. Yet much of the book is taken up with a belabored tracing of the stereo image differences through the many-staged filmic system to the viewer's retinas. Lipton's judgment of his intended audience hobbles the discussion with elementary algebra and proceeds so haltingly as to preclude any comprehensive understanding. Several summary tables are provided, but a straightforward computational scheme suitable for a programmable calculator would have served even better. The general confusion is probably great enough to convince a film-maker that a stereoscopic consultant is a good investment.

Lipton is most interesting when discussing, in a rather anecdotal way. what has worked most effectively in his own films. But despite his worries, it seems that good 3-D movies can't be all that difficult to make. The Soviets have been doing it routinely for decades and have shared their methods through technical publications and demonstrations. Other filmmakers, Felix Bedrossy and Murray Lerner to name two, have worked out their own techniques by dint of observation, perseverance, and talent. The chronic ascendancy of shabby 3-D in the U.S.A. is a frustration to all who savor the richness of visual space, and its roots are only hinted at by Lipton, although with the authority of firsthand experience.

Those who care for language and logic will be dismayed by this hastily produced volume. But anyone with an abiding interest in three-dimensional imaging should be able to justify its place on an otherwise sparsely filled bookshelf.

SR-04C / OPTICAL ENGINEERING / March/April1983 / Vol. 22 No. 2