一种采用提前终止判决法的图像匹配跟踪算法

针对原始遍历式匹配模式算法的时间复杂度高以及给视频图像目标跟踪系统的计算能力带来极大考验的问题,提出一种基于提前终止判决的归一化互相关匹配算法(zero-mean normalized cross correlation based on early termination condition,ZNCC-ETC).在基于盒子滤波优化的遍历式ZNCC匹配算法(ZNCC based on Box-Filtering,ZNCC-BF)的基础上,发现ZNCC-BF算法在遍历匹配过程中存在着一个适当的阈值,一旦模板图与候选图的互相关累加值超过这个阈值,则后续任何计算即为冗余计算,而ZNCC-ETC则采用提前终止判决法实现图像的加速匹配跟踪,并通过对比实验进行验证与分析跟踪结果.分析结果表明:相对于标准ZNCC以及改进的ZNCC-BF,ZNCC-ETC算法能够在保证全局最大值收敛能力的前提下,进一步降低匹配过程中的计算量,实现加速匹配跟踪的目的.

A Matching and Tracking Method Based on ZNCC-ETC

Aiming at high time complexity of original tracking algorithm based on template matching technique and great computations of video image target tracking system, put forwards a matching and tracking method based on zero-mean normalized cross correlation based on early termination condition (ZNCC-ETC). Based on zero-mean normalized cross correlation with box-filtering (ZNCC-BF), there is a proper threshold in the matching process. Once the cross-correlation accumulation value between the template figure and candidate figure exceeds the proper threshold, any subsequent calculation is redundant, however, ZNCC-ETC method is able to realize image speed-up matching tracking by using early termination condition method, and the tracked result is verified and analyzed by comparing the experiments. The analysis results show that, comparing with standard ZNCC and improved ZNCC-BF, ZNCC-ETC can reduce the amount of calculation in the matching process, achieve the goal of speed-up matching and tracking on the premise of guaranteeing the global maximum convergence ability.