Complexity regulation for real-time video encoding

Past research in video compression has focused on improving the rate-distortion performance without active consideration of the encoding complexity. However, realtime software or processor based encoding has tight complexity constraints. In this paper we study real-time software-based video encoding as a problem of dynamic tradeoffs among rate, distortion, and complexity. We present a compression system where the compressed video quality is pre-set and fixed, but the compression complexity is constrained and regulated to ensure realtime performance, by changing the encoding configurations with varying compression efficiency. We propose a dynamic encoding complexity control scheme utilizing an input frame buffer. Our experimental results show that the proposed encoding control algorithm effectively avoids frame drops caused by complexity peaks for real-time encoding.