Multiview 3D Displays: Image Generation Requirements

We are interested in efficient image generation for multiview 3D displays and present results from a recent psychophysical study investigating task performance with different display design parameters. These results have subsequently guided our implementation of novel multiview rendering algorithms. Multiview autostereoscopic displays generate a binocular stimulus to the eyes by producing multiple simultaneous scene views; a single viewer sees just two of these views from any one viewing position. The two key benefits of the multiview approach are; 1) viewers have the freedom to move around and still see a binocular image and 2) there is the ability for several viewers to see a binocular image of the same scene at the same time. Ideally, a multiview display should generate a complete lightfield with full look-around but currently the cost of providing the optical and computational bandwidth is prohibitive. Multiview display designers therefore have to make a tradeoff between: • Number of simultaneous views. • Resolution of each single view. The design choices also depend on the ultimate application, for example whether the display ha