基于数据可重用的运动估计低功耗VLSI实现

Data Reusable Based Motion Estimation Low-power VLSI Realization

块匹配全局搜索算法(FSBM)是效果最好的运动估计算法,但是由于它庞大的运算量限制了它的应用,特别是在数字多媒体终端等对实时性和功耗要求比较高的应用场合。针对这种情况,作者提出的处理结构结合了数据重用和低功耗技术,使得该结构的功耗比传统的全局搜索结构降低50%的功耗,数据可重用性大为提高,确保了每个数据只需被读取一次,这样就确保了较少的I/O端口数和片上内存,同时实现100%的硬件资源利用率。

Block based full search block match （FSBM） is the best motion estimation algorithm, but the huge computation limited its application, especially in the digital multimedia usage situations where need high real time reaction and low power consumption. Confronted with this condition, author brings forward an advanced structure. Some new algorithms are adopted include high data reuse and low power technology. So the new architecture achieves 50% power off and 100% on chip resource usage meanwhile assures low required I/O number and memory.