A fast and robust image matching method for the correlation coefficient

摘要

Convolution computation is one of the key factors by which the efficiency of correlation coefficient matching is restricted. Based on the characteristics of correlation coefficient computation, the bounded partial correlation coefficient matching method is proposed in this paper. For this method, rough matching is carried out with Box-filtering firstly, and then the precise matching is realized by cross search. Finally the sub-pixel precision is obtained by least-square matching. What's more, the correlation coefficient formula is enlarged, and the convolution computation of enlarged portion is omitted, realizing the rapid and robust correlation coefficient matching. Experiments on Unmanned Air Vehicles' image matching have validated the effectiveness and practicability of the method.