More and more people are starting to record their daily lives, whether by traditional video cams or first-person live-loggers attached to glasses, headsets, necklaces or even handbags. Since a group of people (say, at a party) are bound to capture videos of the same event from multiple points of view, Disney Research has developed a method to easily splice together footage taken by different cams. Disney's algorithm chooses the most interesting aspect of a single event based on how many cameras are focused on it, then it chops the videos and chooses parts with best quality, lighting or angle for the final cut.
Since Disney Research touts it as an easy way to create watchable video clips, the system's also capable of following basic cinematographic guidelines without human input, such as avoiding jump shots between cameras. In fact, the team believes it could aid even experienced editors in processing hours upon hours of raw footage. The whole process still takes several hours, but not as long as the average time (20 hours) it usually takes to make a few minutes of edited video.
Wondering about the quality of video clips the algorithm can produce? Disney Research member Ariel Shamir says, "The resulting videos might not have the same narrative or technical complexity that a human editor could achieve, but they capture the essential action and, in our experiments, were often similar in spirit to those produced by professionals." Disney's researchers plan to present the technology at SIGGRAPH 2014 in Canada on August 10th to 14th.
No comments:
Post a Comment