Requirements for Motion Estimation Search Range in MPEG-2 Coded Video

What motion estimation search range is required for inter-frame encoding with the MPEG-2 video compression standard? The answer to this question depends on a number of factors, including video content, video resolution, time-distance between reference and predicted pictures, and just as importantly, pragmatic considerations in implementing a cost-effective solution. For example, in an MPEG-2 encoder chip targeted to MPEG's main level and main profile, the size of the silicon area dedicated to motion estimation circuitry is usually limited by cost, technology, power budget, etc. Limited silicon area translates into a limited number of computations available for estimating motion vectors in each picture of a video sequence. In turn, this computational limitation almost always leads to hierarchical approaches to the estimation problem so as to achieve the search range necessary for high-quality video compression. However, hierarchical methods may also result in sub-optimal motion vector estimation: for a given computational load or silicon area, the larger the search range the less accurate the vector estimates. A designer must thus balance computational limitations with the desire for a large search range and the accuracy of the vector estimation process. In this paper we present a set of experimental results that provide a probabilistic characterization of the size of motion vectors for different types of video: from well-known standard test sequences, to fast-paced sports sequences, to action movie clips. We study the impact of search range on compression efficiency and video quality, and experimentally study the range versus the accuracy tradeoff. Finally, and based on these results, we conclude with recommendations for target search ranges suitable for high-quality compression of standard and high definition video.

Key words: motion, estimation, motion vector