Rate-Distortion-Optimized Video Transmission Using Pyramid Vector Quantization

摘要

Conventional video compression relies on interframe prediction (motion estimation), intraframe prediction, and variable-length entropy encoding to achieve high compression ratios but, as a consequence, produces an encoded bit stream that is inherently sensitive to channel errors. In order to ensure reliable delivery over lossy channels, it is necessary to invoke various additional error detection and correction methods. In contrast, techniques such as pyramid vector quantization (PVQ) have the ability to prevent error propagation through the use of fixed-length codewords. This paper introduces an efficient rate-distortion optimization algorithm for intramode PVQ, which offers similar compression performance to intra H.264/AVC and Motion JPEG 2000, while offering inherent error resilience. The performance of our enhanced codec is evaluated for high-definition content in the context of a realistic (IEEE 802.11n) wireless environment with up to 11 dB PNR improvement over H.264/AVC. We show that PVQ provides greater tolerance to corrupted data while obviating the need for complex encoding tools.