Object's moving speed based model for deciding video sequence's benchmark search range in video coding

Motion estimation consumes most of the computational time in video coding system since the matching between blocks should be carried out on numerous search points (SPs). Adaptive Search Range (ASR) schemes, as one kind of fast ME algorithms, are widely used to reduce the SPs. However, all of the so-called ASR schemes are applied on the precondition that there already has a fixed benchmark SR (bm-SR) exsited. By revealing the correlation between SR and object's moving, this paper proposes a model for deciding the benchmark SR. This proposed model is not only the first work for deciding bm-SR adaptively, but also an adventruous trial for utilizing high level information like object's moving speed.