周泉, 曹辉, 闫博, 杨靓. 高性能图像匹配电路乘累加性能分析[J]. 微电子学与计算机, 2014, 31(8): 56-60.
引用本文: 周泉, 曹辉, 闫博, 杨靓. 高性能图像匹配电路乘累加性能分析[J]. 微电子学与计算机, 2014, 31(8): 56-60.
ZHOU Quan, CAO Hui, YAN Bo, YANG Jing. Performance Analysis of Multiplication and Addition of High-performance Block Matching Circuit[J]. Microelectronics & Computer, 2014, 31(8): 56-60.
Citation: ZHOU Quan, CAO Hui, YAN Bo, YANG Jing. Performance Analysis of Multiplication and Addition of High-performance Block Matching Circuit[J]. Microelectronics & Computer, 2014, 31(8): 56-60.

高性能图像匹配电路乘累加性能分析

Performance Analysis of Multiplication and Addition of High-performance Block Matching Circuit

  • 摘要: 基于归一化互相关的图像匹配在导航制导和模式识别中应用广泛,由于计算量大,应用中通常采用硬件实现.对级联乘累加和并发自累加两种典型结构电路的计算时间性能和资源利用率作了对比分析,从而明确了如何根据具体的匹配尺寸择优选取一种电路完成计算,以获得更好的实时性能和更好的资源利用率,最后从数据复用的角度对存储模型作了说明,并给出了计算核的带宽计算方法.

     

    Abstract: The image matching based on normalized cross-correlation has been widely applied in navigation guidance and pattern recognition,because of the large amount of calculation,the application of the hardware implementation is normally used.There's a comparative analysis between the two typical structural circuits Cascade accumulation and Concurrent accumulation on computing time performance and resource utilization.So as to make clear how to pick over a circuit to complete the calculation according to the concrete matched sizes,which help to obtain a better realtime performance and better resource utilization.At the end,there's an explanation on storage model from the perspective of data reuse.And it gives out the computing method of the bandwidth of computing kernel in the end.

     

/

返回文章
返回