Scalable Video Compression for Embedded Systems in Surveillance

In this paper, we explore coding architectures for camera surveillance applications. For this purpose we propose a Scalable Video Codec (SVC) based on wavelet transformation. The emphasis in the research is on trading-off the coder complexity, e.g. motion-compensation stages, compression efficiency and end-toend delay of the video chain. Within our architecture of choice, t+2D, many complexity-performance trade-offs can be made by configuring the lifting-based motion-compensated temporal filtering (MCTF). We discuss various configurations and elaborate on the required resources for implementation. We have found an SVC (t+2D) that has a scalable complexity and performance, which enables embedded system applications. Even though the complexity of our SVC is many times lower, performance at full resolution is close to H.264 SVC (within 0.2-3 dB depending on the input sequence) and at lower resolutions sufficient for our video surveillance application.