中文摘要：

针对红外图像与全色图像的不同的成像特性，提出了一种基于图像的梯度结构信息的匹配算法．首先对原始的红外和全色图像分别进行梯度计算，得到图像的梯度强度图；对梯度图进行结构相似性度量，获得图像间同名点对．然后采用RANSAC算法剔除误匹配的同名点．最后利用同名点对构建三角网小面元，并进行变换而得到配准图像．实验结果表明，算法可以有效地利用红外图像中地物的结构信息，匹配精度高．

英文摘要：

According to different characters of infrared image and panchromatic image, a new image registration method was proposed based on the gradient structure information of the image. Firstly, the gradient of infrared image and pan- chromatic image were calculated. Then the structure similarity between two gradient images was measured to generate homologous points. RANSAC was used to robustly remove the wrong homologous points. Finally, after dividing image into small regions on triangle irregular network （TIN）, the registered image was obtained by transforming the triangle. Experiment results show that the proposed approach is efficient and the register accuracy is high.