Learned Hardware-in-the-loop Phase Retrieval for Holographic Near-Eye Displays

摘要

Holography is arguably the most promising technology to provide widefield-of-view compact eyeglasses-style near-eye displays for augmented andvirtual reality. However, the image quality of existing holographic displaysis far from that of current generation conventional displays, effectivelymaking today’s holographic display systems impractical. This gap stemspredominantly from the severe deviations in the idealized approximationsof the “unknown” light transport model in a real holographic display, usedfor computing holograms.In this work, we depart from such approximate “ideal” coherent lighttransport models for computing holograms. Instead, we learn the deviationsof the real display from the ideal light transport from the images measured using a display-camera hardware system. After this unknown light propagationis learned, we use it to compensate for severe aberrations in realholographic imagery. The proposed hardware-in-the-loop approach is robustto spatial, temporal and hardware deviations, and improves the image qualityof existing methods qualitatively and quantitatively in SNR and perceptualquality. We validate our approach on a holographic display prototype andshow that the method can fully compensate unknown aberrations and erroneousand non-linear SLM phase delays, without explicitly modeling them.As a result, the proposed method significantly outperforms existing stateof-the-art methods in simulation and experimentation – just by observingcaptured holographic images.