电路板红外图像具有背景区域灰度单调、元器件发热区域轮廓模糊、噪声高、对比度低等特点, 对配准精度和稳定性有很大影响。针对这些特征, 提出了结合小波模极大值和改进 Hausdorff距离的红外图像配准方法。首先用小波模极大值的算法提取边缘曲线; 然后对传统 Hausdorff距离进行改进, 边缘曲线分段, 并建立匹配基本单元; 其次对基本单元粗匹配, 压缩搜索空间, 再进行精匹配, 找出最优变换。最后采用 Matlab进行仿真, 实现图像精确配准。对比实验结果表明: 该方法明显提高了配准精度, 显著改善了配准稳定性。

Circuit board infrared image has characteristics of which the background region gray is monotonous, which the outlines of components’ exothermic areas are rough, high noise and low contrast. And these features have a great influence on the registration accuracy and efficiency. According to these characteristics, this paper proposed an infrared image registration method combining wavelet modulus maxima and the improved Hausdorff distance. First, the edge curves were extracted by using wavelet modulus maximum algorithm, then improved the traditional Hausdorff distance, conducted curves segmentation and established basic unit. Secondly, rough match the basic unit, compress the search space, then conduct fine matching to find out the optimal transform. Finally, use MATLAB simulation to achieve precise image registration. Experimental results show that: this method can obviously improve the registration accuracy and significantly improve the stability of the registration.