POWER EFFICIENT MOTION ESTIMATION USING MULTIPLE IMPRECISE METRIC COMPUTATIONS

In this paper, we propose power efficient motion estimation (ME) us-ing multiple imprecise sum absolute difference (SAD) metric com-putations. We extend recent work in [18] by providing analytical solutions based on modelling of computation errors due to voltage over scaling (VOS) and sub-sampling (SS). Results show that our so-lutions provide significantly better performance in the sense of rate increase for fixed QP, e.g., less than 5 % increase, while in [18] the rate increase could be as high as 20%. Our analysis also allows us to compare different ME algorithms (e.g., full search vs. a fast al-gorithm) and SAD computation architectures (parallel vs. serial) in terms of their robustness to imprecise metric computations and their power efficiency. Finally, we demonstrate that additional power sav-ings can be achieved by removing redundancy between the various computations. Index Terms — voltage over scaling (VOS), error tolerance (ET), matching metric computation (MMC), imprecise computation 1