A data reusing architecture for MPEG video coding
Vasily G. Moshnyaga, K. Masunaga, Naoki Kajiwara · 2004
This paper presents a new architecture capable of drastically reducing computations and memory accesses in MPEG video coding. The key idea is to reuse data stored in frame memory to disable memory re-writes and motion estimation computations for stationary macroblocks. Due to dynamic memory sharing among reference frames, data-driven macroblocks characterization and processing, the architecture reduces the number of frame memory writes and motion estimation operations by 85% while maintaining high picture quality.