Affiliations: College of Mechanical and Electronic Engineering, Northeast Forestry University, Harbin, China
Correspondence:
[*]
Corresponding author. Peng Wu, College of Mechanical and Electronic Engineering, Northeast Forestry University, Harbin 150040, China. E-mail: wupeng@nefu.edu.cn.
Abstract: Normalized cross-correlation (NCC) is fast to compute but its accuracy is low. In this paper, we propose a fast, highly accurate NCC image matching algorithm. First, a wavelet pyramid is constructed to reduce feature point searching and matching times. Then, an NCC image matching algorithm is used to acquire the coarse matching points in the original image. Next, an improved iterative relaxation algorithm is used to remove false matching points, and an adaptive winner-takes-all strategy is introduced to improve the algorithm’s iteration speed and obtain more one-to-one matching points. Our experimental results show that the proposed algorithm can improve not only matching speed, but also matching accuracy.
