A Low-Power Video Decoder with Power, Memory, Bandwidth and Quality Scalability

摘要

This paper describes a low-power scalable video decoder for use in portable video applications. The scalable video decoder uses tree structured vector quantization (TSVQ) of perceptually weighted block transforms. The subjective quality of compressed images improves significantly by the use of perceptual distortion measures. The low-complexity, low-power architecture requires only table-lookups to perform video decompression. Inverse transforms are performed as pre-processing steps in the tables. Color conversion from YUV to RGB and color quantization are also performed as pre-processing steps in the tables. The video decoder provides a trade-off between rate-distortion, power and memory size. This allows to trade-off power and memory size for better quality of compressed video and vice-versa. The power consumption of our video decoder is orders of magnitude smaller than other decoders in existing technology. Measured performance shows that the scalable video decoder consumes between SO to 150 micro-watt with a 1.5 V power supply in 0.8μ CMOS technology for 160×240 resolution video at 30 frames per second.