Scene context dependent key frame selection in streaming

Abstract

In this paper, we describe the development of our scene context dependent key frame selection method to reduce the amount of recorded video data. We propose the use of motion analysis (MA) to adapt to scene content in our key frame selection process. Based on the information generated by the motion analysis stage, frames in the video sequence which contain significant motion information are selected to retain for recording. We also show that our proposed method performs better than the traditional time-lapse recording method.