Motion Parallax in Stereo 3D: Model and Applications

摘要

Binocular disparity is the main depth cue that makes stereoscopicrnimages appear 3D. However, in many scenarios, the range of depthrnthat can be reproduced by this cue is greatly limited and typicallyrnfixed due to constraints imposed by displays. For example, due tornthe low angular resolution of current automultiscopic screens, theyrncan only reproduce a shallow depth range. In this work, we studyrnthe motion parallax cue, which is a relatively strong depth cue, andrncan be freely reproduced even on a 2D screen without any limits.rnWe exploit the fact that in many practical scenarios, motion parallaxrnprovides sufficiently strong depth information that the presence ofrnbinocular depth cues can be reduced through aggressive disparityrncompression. To assess the strength of the effect we conduct psychovisualrnexperiments that measure the influence of motion parallax onrndepth perception and relate it to the depth resulting from binocularrndisparity. Based on the measurements, we propose a joint disparityparallaxrncomputational model that predicts apparent depth resultingrnfrom both cues. We demonstrate how this model can be applied inrnthe context of stereo and multiscopic image processing, and proposernnew disparity manipulation techniques, which first quantify depthrnobtained from motion parallax, and then adjust binocular disparityrninformation accordingly. This allows us to manipulate the disparityrnsignal according to the strength of motion parallax to improve thernoverall depth reproduction. This technique is validated in additionalrnexperiments.