Design and Analysis of Motion Estimation Architecture Applicable to Low-power Energy Management Algorithm

저전력 에너지 관리 알고리즘 적용을 위한 하드웨어 움직임 추정기 구조 설계 및 특성 분석

The motion estimation which requires huge computation consumes large power in a video encoder. Although a number of fast-search algorithms are proposed to reduce the power consumption, the smaller the computation, the worse the performance they have. In this paper, we propose an architecture that a low energy management scheme can be applied with several fast-search algorithm. In addition. we show that ECVH, a software scheduling scheme which dynamically changes the search algorithm, the operating frequency, and the supply voltage using the remaining slack time within given power-budget, can be applied to the architecture, and show that the power consumption can be reduced.