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.