Gradient-based 2D-to-3D Conversion for Soccer Videos

Kiana Calagari, Mohamed Elgharib, Piotr Didyk, Alexandre Kaspar, Wojciech Matusik, Mohamed Hefeeda
2015 Proceedings of the 23rd ACM international conference on Multimedia - MM '15  
A wide spread adoption of 3D stereoscopic television is hindered by the lack of high-quality 3D content. One promising solution to address this need is to use automated 2D-to-3D conversion. However, current conversion methods, while general, produce low-quality results that exhibit artifacts that are not acceptable to many viewers. We address this problem by showing how to construct a high-quality, domain-specific conversion method for soccer videos. We propose a novel, data-driven method that
more » ... enerates stereoscopic frames by transferring depth information from a database of similar 3D stereoscopic videos. While a database of 3D stereoscopic video with accurate depth is very difficult to obtain, our key observation is to leverage high-quality computer generated content in current sport computer games to generate a reference data set. Once we retrieve similar 3D video frames, our technique transfers depth gradients to the target frame while respecting object boundaries, solves for the output depth map, and then generates stereoscopic video. We validate our method by conducting user-studies that evaluate depth perception and visual comfort of the converted 3D videos. We also show that our method significantly outperforms the current state-of-the-art method. For example, up to 20% improvement in the perceived depth is achieved by our method, which translates to improving the mean opinion score from Good to Excellent.
doi:10.1145/2733373.2806262 dblp:conf/mm/CalagariEDKMH15 fatcat:gosmwahnnncjjf6b5643adm3em