针对画幅式扫描成像方式获取的红外图像，传统的图像拼接技术无法有效地完成红外图像的实时拼接。结合基于图像点特征算子的实时拼接技术，对前沿的点特征提取及描述算子进行了模糊、噪声、像移、旋转下的性能研究，结果表明ORB 特征提取算子及BRISK 特征描述算子的性能最符合画幅式红外拼接系统的需要。最后在CPU＋GPGPU 平台上实现了以ORB＋BRISK 算子为特征算子的红外图像实时拼接系统，并利用热红外遥感图像进行了拼接的性能验证。

As for the infrared image acquired by frame sweeping, traditional methods of image mosaicking cannot effectively accomplish the real-time infrared image mosaicking. The paper combines with the image mosaicking based on point feature algorithms and compares the performance of latest point feature detectors and descriptors under conditions of Gaussian blur, Gaussian noise, motion blur and rotation. The result shows ORB detector and BRISK descriptor are most consistent with frame infrared image mosaicking. The paper finally implements the real-time mosaicking system of frame infrared image based on ORB＋BRISK feature algorithm on CPU＋GPGPU platform, and verifies the mosaicking performance using infrared remote-sensing images.