Data-Driven Techniques for Low-Energy Video Processing

The explosive growth of portable video applications has elevated needs for low-energy video processing hardware. This paper presents algorithmic and architectural techniques capable of reducing the energy dissipation of video codec considerably. The focus is put on adaptive approaches, which dynamically adjust the number of operations, the operational bit-width, and voltage supply to varying picture content. We describe the strengths of techniques in terms of energy savings and design complexity and overview issues of their practical use.