Kinder Gentler Stereo

Abstract

It is of course well known that many sensory phenomena, both binocular and monocular, contribute to the human sensation of depth. Discussion of binocular modalities is generally restricted to perspective disparity, the left eye / right eye differential in the apparent relative position of objects in the scene depending on their depth; however it has been recognized for many years that other sources of binocular disparity, e.g., differential distortion between two renderings of the same scene from the same perspective, also stimulate the depth sensation [1]. Monocular depth perception modalities based on more-or-less high level understanding of real world structures and motion dynamics are also well known; however several authors have also speculated about the possibility of one or more low level monocular depth perception modalities, e.g., making use of the very small parallax differentials due to eyeball tremor. We look systematically at the synergistic combination of these phenomena in an attempt to deliver "kinder gentler stereo (KGS)", a stereo capture, rendering, and display concept that (1) evades cue conflicts, e.g., the well known convergence-accommodation conflict, that contribute to difficulty in fusing stereoscopic image pairs, and the (perhaps subliminal) viewing discomfort of seeing even successfully fused images in the presence of residual conflicts with other depth perception modalities, and (2) avoids the technical and social impediments of 3D-display systems that require the user to wear eyeglasses, hold his or her head in a particular location with respect to the screen, preclude multiple viewers of the same screen, etc. Kinder gentler stereo should require zero "eyeball lock in" time [2] when first presented, look normal (i.e., either correctly stereoscopic or apparently monoscopic) from any position from which the screen is visible, look normal monoscopically (e.g., to a one-eyed viewer), and deliver a measure of binocular stereopsis commensurate and consistent with other depth sensation stimulating components of the imagery. Toward this end, we investigate optical and digital image processing steps that, while they introduce distortions that are contrary to strict "geometrical correctness" [3], generate increased rather than decreased stereoscopic viewing comfort. We examine the lower limits of interocular separation (or its equivalent in, e.g., so-called "single lens stereo"), finding that a surprisingly small fraction of the nominal 65 mm interocular equivalent separation can stimulate accurate and extremely pleasant depth sensations. Via these adjustments, and by appropriately selecting suitable image capture, rendering, and display parameters, we have delivered imagery that is not apparently stereoscopic (doubled, ghosted) when viewed without stereoscopic eyewear. This suggests that "microstereoscopic" imagery might be viewable stereoscopically by means of a multiplexing system that admits substantially more crosstalk than is tolerable in conventional systems; preliminary experiments support this hypothesis. Finally, the possiblility is suggested that, based on this observation, a zoneless autostereoscopic stereoscopic display system may be feasible; current work focuses on pursuing this possibility. [1] Alfred H Schwartz, Stereoscopic Perception with Single Pictures, Optical Spectra, 1971-Sep p25-27. [2] Time required to coordinate convergence and accommodation and perceive a stereo pair as a single scene with depth rather than as doubled vision. [3] V. S. Grinberg, G. W. Podnar, M. W. Siegel, "Geometry of Binocular Imaging", Stereoscopic Displays and Applications V, 1994, 56-65.