Optimal Video Summarization with a Bit Rate Constraint

Video summarization offers an intelligent way of reducing the video content temporally to satisfy the user preference for a shorter viewing time, and/or to meet the bit budget limitations in storage/communications. In this paper we formulate the summarization problem as a rate-distortion optimization problem under bit budget and frame skip constraints, and introduce a summarization distortion metric that is based on the average frame distortion caused by dropped frames. An optimal summarization solution is therefore developed based on Lagrangian relaxation and dynamic programming. The simulation results demonstrate the effectiveness of the proposed summarization technique.