一种改进的快速归一化积相关图像匹配算法

归一化积相关算法是一种经典的图像匹配算法,具有操作简单、匹配概率高等优点,其不足之处主要在于计算量大,难以满足实时性要求。为此,本文首先提出了一种减少归一化积相关算法计算量的方法,通过从基准图像中构造出两个搜索矩阵,来简化实时图像的搜索路径,使其只需沿一个方向平移就可以完成匹配,并且可以很容易地通过相邻两个基准子图之间的迭代来避免图像能量的重复计算;接着将该方法与BPC算法的思想相结合提出本文的快速匹配算法,从两个方面来减少传统算法的计算量,进一步提高了算法的运算速度;最后通过实验仿真验证了本文算法的优越性。

An Improved Fast Normalized Cross Correlation Algorithm for Image Matching

The normalized cross correlation algorithm is a classic image matching algorithm. The method is simple and has a high matching probability. However, its disadvantage is enormous computation, which is difficult to meet the requirements of real-time. In this paper, first, a method for reducing the computation is proposed. By constructing two searching matrix from the reference image, the searching path of the real-time image is simplified. It completes the matching just along one translation direction and can avoid repetitious computation of the image energy easily based on an iterative operation between the two adjacent sub-based images. Then, the fast matching algorithm is proposed by combining the method and the bounded partial correlation (BPC) algorithm. It improves the computational speed of the algorithm by reducing computation from two aspects. Final, the experiment results show the superiorities of the algorithm.