Motion estimation algorithm for scalable hardware implementation

Motion estimation is an indispensable stage of motion pictures encoding used in all common video coding standards, such as MPEG-2 or AVC. At the same time it is the most complex operation both in software and hardware implementation of an encoder that makes the choice of the appropriate motion estimation algorithm crucial in terms of increasing the entire coding process efficiency. This should be taken into consideration especially in case of real time processing. In this paper, a novel scalable multilevel algorithm for motion estimation is proposed. It is a fast motion estimation algorithm, using the block-matching mechanism, provided the motion vectors accuracy equal frac14 of the sampling period. Proposed algorithm is modular and scalable. First feature enables structure scalability, second provides mechanism to control of clock cycles number spend for motion estimation process. As a result the global clock-rate required for real-time processing can be adjusted.