Journal of the Society of Motion Picture Engineers (1930-1949)

conform to the procedures of the flat film, yet without loss of effect. In this Part we shall discuss in some detail, using the method of analysis already derived from the general theory, two typical proposals, both in current use and both claimed to provide a perfect solution to all problems of stereoscopic transmission. The first makes use of a fixed interaxial separation of the camera lenses, and a variable convergence; in the second the convergence is fixed, but the interaxial separation is made variable. Viewing of Real Objects and Stereo Images It has been suggested — notably by Dewhurst9 in Great Britain and more recently by an influential group in the U.S. — that the problem of 3-D filming is very simple: all that is necessary is to provide a fixed lens separation (tc) approximating that of the human eyes and then converge the optical systems on some appropriate plane in the scene — this plane appearing, of course, in the plane of the screen when the film is projected with ZCL = 0. It has been further suggested that the convergence control ought to be coupled to the lensfocusing mechanism in such a way that DI is always the distance to the plane of sharpest focus. Thus, following focus would automatically alter the convergence, and (since tc is already fixed) no special stereoscopic adjustments of any kind would require to be made. This, it is held, would reproduce the conditions of natural vision, and would provide strain-free viewing of the image by all spectators.* * The same idea has recently occurred to the first producer of 3-D motion pictures in Hungary, M. Felix Bodrossy. "The Hungarian method," he writes, "solves the problem simply and radically: it starts off from the way the eyes work, and imitates nature. The eyes always focus automatically and at the same time converge on the object they look at. Our cameras do the same thing."10 The parallel with the human eyes is simple and attractive; but, from what has been said above, it will be apparent that the viewing of a stereoscopic image cannot at present be made to resemble human vision at all closely. The image in space is not even an optical image; it is a mental construction from data supplied solely by overlapped images on a flat screen. This construction is accomplished by methods not used in normal vision; for example, the spectator's eyes must remain focused at the screen distance, but they will be varyingly converged according to the distance of the point of attention, which may be much nearer or much more remote. Furthermore, in the real world, sense-data remain more or less constant when spectator and scene are fixed; but stereoscopic data may be made to vary widely according to projection conditions, and indeed cannot be kept constant when the size of the screen is changed. It is therefore not to be expected that a mere reproduction at the camera of the human eye separation — in the absence of human viewing methods — will of itself produce strain-free viewing. This cannot be so simply achieved until it becomes possible to create real or virtual 3-D images in space. Limitations of Fixed-fc Systems Meanwhile, stereo camera systems which make use of a fixed "human" lens separation of 2.5 in. must be treated as having an awkward limitation common to all fixed lens systems designed to film pictures for large screens. The transmission system obtaining with a fixed value of tc can be very clearly exhibited on a graph similar to Figs. 8 and 9. Reference to the section "Depth Range in the Scene," earlier in this paper, will show that the slope of the transmission lines is a function of the C factor (i.e. Mfctc) and t; and therefore, if M is assumed fixed for the film, tc is fixed on principle, and fc represents the focal length of the lens in use, all possible trans 276 October 1952 Journal of the SMPTE Vol. 59