16.1 INTRODUCTION
Motion estimation plays a key role in several applications such as video on demand, high-definition TV (HDTV), and multimedia communications. Motion estimation is used to remove temporal data redundancy between successive video frames. Video data compression rate can be improved by estimating the offset information of objects from one frame to another (motion estimation) and then by encoding only frame differences with respect to that offset (motion compensation).
There are several types of motion estimation: pixel based, block based, and region based, where motion estimation can be done pixel by pixel, block by block, or region by region, respectively. Block-based motion estimation is the most popular method due to its simplicity and suitability for hardware implementation. Block-based motion estimation is used in MPEG, H.263, H.264/AVC, and other video compression standards that aim at achieving high video compression ratios at real-time speeds, which requires a huge amount of computations. As a result, any efficient hardware would require the use of single instruction multiple data stream (SIMD) processors, which are special-purpose very large-scale integration (VLSI) circuitry employing a high degree of parallelism while requiring little input/output (I/O) communications. The main design challenge focuses on how to arrive at SIMD processor architectures that satisfy system-level requirements for a given complex algorithm with maximum processor utilization ...
Get Algorithms and Parallel Computing now with the O’Reilly learning platform.
O’Reilly members experience books, live events, courses curated by job role, and more from O’Reilly and nearly 200 top publishers.
