Optimized single-viewer mode of multiview autostereoscopic display

We propose an approach for optimizing the visual quality of a multiview 3D display for a single viewer. The approach combines eye-position tracking system with on-the-fly visual optimization of multiview image content. The tracking algorithm uses the video input from a pair of off-the-shelf web-cameras and employs fast and robust face and facial feature detection algorithms to provide features for the subsequent stereo matching and distance estimation. Based on display measurements and having user's eyes position, the following visual improvements are achieved: continuous head parallax for wide range of observation angles, cross-talk mitigation and brightness enhancement for a single viewer, and multi-view image compensation related to the distance of the observer.