Chroma Prediction for Low-Complexity Distributed Video Encoding

Inter-frame encoding is an integral technology in modern video codec design, used to achieve high compression efficiency by reducing temporal redundancy. An important step in inter-frame encoding is the temporal prediction loop, involving computationally expensive block-matching algorithms. When the video server is concerned with customer Quality of Service for on demand or live-streaming, the encoder is the appropriate location for most of the compression computation to enable low client decoding time. However, emerging use-cases such as Wireless Video Sensor Networks (WVSNs) and remote surveillance present new technical challenges in compression. For these scenarios, the video capture device may be power-constrained with limited resources, and it is better that the computational burden fall on resource rich decoders/transcoders available at data centres. In this paper, we present a video coding scheme with low encoding complexity that passes the computational burden to the decoder. Bit-rate reduction is achieved by removing Chroma information from specific frames and then re-colouring them at the decoder using a technique called Motion Compensated Recolouring (MCR). MCR can reduce bit-rate by as much 16% compared to Intra-only video coding, with only a slight drop in objective quality. We also demonstrate that MCR out-performs other Chroma prediction techniques at higher quality and for videos with complex motion.