| Abstract: | Taking into consideration computational complexity and design regularity, this article proposes a content‐aware full search (CAFS) block matching scheme for motion estimation. Full search (FS) is widely employed in hardware design of block matching because of its regular data flow. But the huge computational complexity of FS makes it infeasible especially in low‐power environments. To reduce the required computational complexity, CAFS using a content‐aware computation allocation mechanism dynamically distributes computation to blocks with distinct motion content based on center‐biased distribution characteristics of motion vectors. Because all operations of CAFS can be performed by a series of ±2 FSs, the hardware of CAFS can be easily extended by the design of FS. Experimental results indicates that CAFS can achieve about 0.5–1.4 dB quality improvement over original FS in football and table tennis sequences under the same computation. © 2005 Wiley Periodicals, Inc. Int J Imaging Syst Technol, 14, 246–252, 2004; Published online in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/ima.20028 |
