Efficient View Synthesis from Uncalibrated Stereo

摘要

For multiview auto-stereoscopic 3D displays, available stereo content needs to be converted to multiview content. In this paper we present a method to efficiently synthesize new views based on the two existing views from the stereo input. This method can be implemented in real-time and is also capable of handling uncalibrated stereo input. Good performance is shown compared to state-of-the-art disparity estimation algorithms and view rendering methods.