Abstract

Novel volumetric multi-focal displays can reproduce all desired depth cues while maintaining a high-quality resolution and colour gamut. Therefore, they show a potential to revolutionize the current design of head-mounted VR/AR displays, which lack accommodation depth cues. Rendering a high-quality scene to a multi-focal display involves a non-trivial optimization process. Unfortunately, even after intensive optimizations, current multi-focal rendering algorithms do not generate a scene that creates a satisfactory impression of realism. This is because these optimizations are based on a simplistic loss which fails to take into account the complexities of human visual perception as a whole. The objective of this proposal is to employ adaptive methods to efficiently identify the most relevant visual cues that contribute to an impression of realism, devise a holistic full-reference 3D-scene quality metric, and design a new optimization-based multi-focal rendering algorithm based on the metric.