Practical Temporal Psychovisual Modulation with Liquid Crystal Devices

Abstract

Temporal Psychovisual Modulation (TPVM) is a new paradigm of computational display, which can concurrently exhibit many different views on a common display. These views are decomposed into a set of atom frames, and the atom frames are modulated by users' liquid crystal (LC) light modulation glasses. Due to the limited refresh rate of LC glasses, a practical TPVM multiview display system can only support two to three atom frames. Such a small number of atom frames are not sufficient to a multiview application with a high image quality. Therefore, the main technical challenge before TPVM is how to support as many viewers as possible while maintaining an acceptable perceptual quality, using only a small number of atom frames. In this thesis, we develop two approaches to meet the challenge.

The first approach is to exploit the sparsity of multiview images to be displayed. One example is the sequence of depth of field (DOF) images. Those images not only provide a strong depth cue, but also have a sparsity structure. That structure allows the DOF images to be reconstructed from a small number of atom frames. We prove that property theoretically. Experimental results also agree well with the proof.

The second approach is to exploit the well-known property of rapidly decreasing visual acuity from fovea to peripheral vision. The strategy is to exhibit different concurrent views at highest quality in viewers' focused regions, while allowing graceful image quality degradation in regions of peripheral vision. This is achieved by a novel fovea weighting algorithm that optimizes for subjective quality. We find the proposed algorithm improves viewers' perceptual quality significantly, especially when the TPVM multiview display system only has a small number of atom frames.