Information reduction based on temporal similarity and spatial importance for video transmission in mobile surveillance system

In video surveillance system, a large number of video measurements would increase network load and degrade control performance, and make data storage a difficult task. Hence, how to adaptively reduce the data quantity and preserve the essential information of video data becomes an important research issue. In this paper, an adaptive adjustment based on temporal similarity and spatial importance is proposed to effectively eliminate temporal redundancy while maintaining high importance content from video data. By analyzing video data, it is observed that temporal similarity can be easily found in sequential video images and, on each image frame, different levels of importance can be further characterized in the spatial domain. Moreover, the bandwidth allocated to transmit these processed video data can be dynamically adjusted based on user demand. Finally, two scenarios of experimental tests are extensively evaluated and experimental results demonstrate the exceptional performance of the proposed algorithm.