NC State University Visual Experience Lab
The latest light-field displays (LFDs) have improved greatly with head tracking, but continue to be based on the approximate pinhole model, meaning that cross-talk is often visible at large viewing angles, and when virtual objects are distant from the display panel. To address these problems, our real-time LFD technique evaluates a full optical model for every frame, and then displays an image predistorted during ray-traced rendering at the sub-pixel level to the current pixel-to-eye light flow, reducing cross-talk and increasing viewing angle. A comparison to imagery produced with the pinhole model shows clear advantages for our LFD technique; and in a user study with interactive, head-tracked display of several different scenes, viewers reported a significant preference for our novel LFD technique. Our unoptimized prototype implementation is already real-time, and as GPUs improve in performance and ray tracing becomes increasingly common, we expect our LFD technique will be easy to add to a variety of 3D applications, including games. Our technique is easy to adapt to any current LFD using lenticular lenses.