Reducing Power Consumption of Block Matching Motion Estimation Using Adaptive Algorithm Selection

Full Search FS is regarded as optimal for block matching motion estimation algorithm with high power consumption while fast search such as Three Step Search TSS produces sub optimal motion vectors with a lower power consumption Real time adaptive selection between these two algorithms can provide a good trade o between picture quality and power and reduce the power consumption while main taining high picture quality In this paper a real time algorithm selection scheme is proposed which adaptively replaces FS by TSS for some macroblocks in order to reduce power consumption Simulation results on MPEG video sequences show that over power reduction can be achieved by using dynamic thresholds without signi cant quality degradation in comparison to Full Search only motion estimation