摘要 |
A method, apparatus and computer program product are provided for generating semantic information from video content. Objects and regions of interest within video content may be identified and monitored for characteristics relating to object detection, motion content, and motion trajectory. Salient events relating to the regions may be detected based on the monitoring. Temporal segments may be identified and used to create summary video content, or highlights. An example embodiment relates to processing video footage of sports. Goals, scored points, unsuccessful scoring attempts, as well as other events may be detected in the video content. Efficiency is gained by monitoring only a relatively small portion of the frame, and by limiting the dependency on tracking moving objects. |