Data Reduction Based on Keyframe with Motion Energy Extraction Rules

For improving the public safety, upgrading the visual experience in entertainment and sports, and assisting teaching in education, the demands for video-related information are rapidly increasing. Video packet data are transmitted to the controller or end-users for further analyzing and enjoying. However, bandwidth of channel and computation is limited. Over required data would cause congestion to loss certain important video shot or frames to influence performance. In order to solve the problem, data reduction is necessary. However, over data reduction would make system performance become worse. Hence, for achieving the purposes of reduction and performance guarantee, keyframe extraction rules based on motion energy is proposed. Four tested videos are used to demonstrate the efficiency and intelligent extraction result and more important is system performance is kept in an acceptable range or even better than origin one. Furthermore, other two experimental videos are utilized to present the comparison results of traditional, the proposed extraction rules, fixed interval sampling and triangle-based. The experimental and comparison results demonstrate the outstanding performance of the proposed extraction method which only uses 50% video data.