基于特征点自动匹配的红外图像配准研究

在模糊和含噪声的红外图像配准中,利用角点检测实现特征点的选择,在提高角点提取效率的同时又保证了角点提取的精度。根据互相关的双向匹配实现对应特征点的自动匹配,然后由对应的特征点对估计出仿射变换的参数。实测的数据和计算结果表明,这种方法对于双波段红外图像的配准是有效的,而且有利于后续的图像融合。

Dual-Band IR Images Registration Based on Feature Points Auto-Matching

A high performance feature point selection algorithm with corner detection is presented to handle blurred and noisy infrared images registration. The detector improves efficiency and the accuracy of corner detection. Bidirectional cross-correlation is used for feature points auto-matching and the affine transformation parameters are estimated from the correct corre sponding points. The experimental resuhs demonstrates that this method is effective for dual-band IR images registration and beneficial to image fusion.