基于哈特莱变换的快速图像模板匹配算法

Fast Image Template Matching Algorithm Based on Discrete Hartley Transform

去均值归一化模板互相关(ZNCC)是工程中应用最多的图像匹配算法,但过高的计算复杂度严重限制了其在实时系统中的应用。针对这一问题,提出了基于快速哈特莱变换的快速模板图像匹配算法,首先推导了该算法在哈特莱域的表达式,利用可分离的快速哈特莱变换对相关面进行高效率整体计算,然后在空间域对获取的相关面进行快速归一化处理和极值搜索,并通过空间换取时间和积分图的策略进一步加快算法的计算速度。对算法计算量的定量分析和仿真实验结果表明,算法计算效率高,并且可以完全重构,加速比与图像内容无关,综合性能全面优于现有算法,具有良好的工程应用前景。

The exhaustive-search zero-mean normalized cross-correlation algorithm（ZNCC） is one of the most widely used image matching algorithms in the autonomic navigation systems of unmanned aerial vehicle.However,the computational cost of the algorithm is too high for many real-time applications.A fast image template matching algorithm based on discrete Hartley transform is proposed to greatly reduce the computational complexity of the algorithm.First,the ZNCC algorithm is described in the Hartley domain.Second,the cross-correlation surface is obtained by the separable 2D fast Hartley transform.Third,the inverse Hartley transform is calculated,and then the normalization and extremum search are conducted in the space domain.Finally,the computational complexity is further reduced by the image-integration strategy.The effectiveness of the proposed algorithm is demonstrated by using the SAR/optics matching tasks.Experimental results show that the proposed algorithm has high precision,high computational efficiency,data independence,and it outperforms the other popular state-of-the-art fast algorithms and would be very suitable for the image-aided navigation applications of unmanned aircraft.