Joint Coding-Transmission Optimization for a Video Surveillance System With Multiple Cameras

Abstract

Surveillance video has become an important multi-media application in recent years, with development trends toward wide viewing angles and high definition. At the same time, transmission of surveillance video over the Internet and storage in the cloud are becoming increasingly popular. However, it is extremely challenging to transfer surveillance video with both a wide viewing angle and a high resolution over the Internet because of 1) the tradeoff between a wide viewing angle and high-surveillance object resolution for a single camera, 2) the large throughput requirement, and 3) the fluctuating delay and bandwidth of the Internet. In this paper, we present a reference-frame-cache-based surveillance video transmission system (RSVTS), which delivers wide-viewing-angle and high-definition surveillance video over the Internet in real time using multiple rotatable cameras. First, we develop an efficient video mosaicing method for merging two (or more) surveillance videos together. Second, novel reference frame selection and update algorithms are proposed for the RSVTS encoder to reduce the amount of encoded data. Third, we implement a reference frame cache on both the sender and receiver sides to increase video quality by increasing the probability of video decoding. We conduct a performance evaluation using a simulation system that integrates RSVTS and ns-3. Compared with the H.264/SVC and H.264/AVC schemes, RSVTS achieves appreciable improvements in enhancing the video peak signal-to-noise ratio and bandwidth conservation.