Jing Liu, Tom Malzbender, Siyang Qin, Bipeng Zhang, Che-An Wu, James Davis
12/18/2014 12:18 PM
Computer Science
Multiview autostereoscopic displays have several image artifacts which prevent widespread adoption. Crosstalk between adjacent views is often severe, stereo inversion occurs at some head positions, and legacy 2-view content is difficult to display correctly. We introduce a method for driving multiview displays, dynamically assigning views to hardware display zones, based on potentially multiple observer’s current head positions. Rather than using a static one-to-one mapping of views to zones, the mapping is updated in real time, with some views replicated on multiple zones, and some zones left blank. Quantitative and visual evaluation demonstrates that this method substantially reduces crosstalk.
