一种改进的红外图像归一化互相关匹配算法

An Improved Normalized Cross-correlation for Template Matching of Infrared Image

分析了传统归一化互相关算法在红外空中目标匹配定位时失效的原因,提出一种改进的红外图像归一化互相关匹配算法.该方法将模板和匹配区域之间的纹理相关计算看作一个最优化问题,寻求使图像纹理相关匹配鲁棒性最好的相关基准值,用图像的相关基准函数替代传统方法中的区域均值部分,构造了一种适用于的红外目标匹配的归一化相关算法.实验结果表明,该相关匹配算法对模板中背景部分的变化和非均匀性亮度变化有良好的抗干扰能力,较好地解决了恶劣环境下红外对空目标跟踪中匹配定位出错的问题.

A novel normalized cross-correlation paid （NCC） method for template matching of infrared image was proposed. Although the classical NCC method paid attention to global correlation for template matching,it ignored the correlation of rows and columns texture between template and image regions. Therefore,the classical NCC method might fail for template matching in complex scene. To improve its performance, the NCC computation was regarded as an optimization problem, which aimed to make algorithm most robust. And then the average in classical NCC formula was replaced by the optimization function of NCC reference value. So a novel NCC algorithm was brought forward,which was well suited for template matching of infrared image. The experimental results over real-world sequences show that the novel NCC algorithm is better than the classical algorithm in complex scene and is robust to the changes of background and target in template.