Summarizing Surveillance Video by Visual Saliency and Moving Object Information

Everyday an enormous amount of video is captured by surveillance system for various purposes around the whole world. However, this is almost impossible for human to analyze the vast majority of video data. In this paper, a video summarization method is introduced combining foreground object, motion, and visual attention cue. Foreground objects typically provide important information about video contents. Additionally, object motion is naturally more attractive to human being. Moreover, visual attention cue indicates the human’s attraction label for key frame determination. Using these features, supervised classifier support vector machine (SVM) is applied to obtain the key frames from a surveillance video. Extensive experimental results show that the proposed method performs superior to the state-of-the-art method using publicly available surveillance video dataset BL-7F. Keywords—Background modelling, foreground object, motion information, graph based visual saliency, support vector machine