Low power HD video fast motion estimation algorithm based on signatures

Abstract

Video motion estimation consumes the major part of time and power in any video compression standard. Today most video capturing devices record HD video and are battery operated, which creates the need for low power Fast Motion Estimation algorithms This paper presents such an algorithm, which targets Full Search quality at HD resolution. It is built upon existing Fast Motion Estimation algorithms for lower resolution and introduces the concept of signature based motion estimation. The proposed algorithm hashes the Low Pass filtered video information to quantify Similarity. The proposed algorithm succeeds in closing about half of the quality gap between some of the most efficient Fast Motion Estimation algorithms and Full Search.