关键词: 红外, 可见光, 图像配准, 二阶互信息, 对齐度, Hough变换

Abstract:

In this paper,a registration algorithm that combines second-order mutual information with alignment as a similarity function is proposed.Firstly,the consistency lines are extracted from visible/IR images by using Hough transform.Secondly the alignment metric of consistency lines is calculated.Finally,these different images are registered by adjusting the proportion of alignment metric in the new similarity function.Experiment results demonstrate that the proposed similarity function shows higher accuracy and robustness than the conventional similarity functions.

Key words: IR;visible;image registration;second-order mutual information;alignment metric;Hough transform