红外视频的SIFT-RANSAC稳像算法研究

针对红外热像仪在载体的非平稳运动中获得的视频信号存在帧间的非平稳抖动问题，本文提出了一种SIFT—RANSAC稳像方法，将图像匹配中的SIFF特征应用于红外视频的稳像处理中，并针对红外热像仪的成像特点，给出了一种基于RANSAC的方法估计运动参数，对运动参数的滤波得出无抖动的运动轨迹，再对每帧图像做运动补偿得到无抖动的视频。仿真结果表明，该方法有效地解决了特征点匹配过程中的误匹配问题，大幅度地提高了红外视频的稳像精度。

Research on SIFT-RANSAC Stabilization Algorithm of Infrared Video

The video got from infrared thermal imager should be stabilized to remove annoying shaky motion caused by unwanted motion of the platform. Aiming at the problem, according to the characters of the infrared thermal imager, a stabilization algorithm of SIPF-RANSAC is proposed for motion parameters estimation with SIFT features and RANSAC method. A motion filter is used to obtain a motion curve without shaky noises. A stabilized video is gained by motion correction and compensation. The result proves that the algorithm is precise enough for infrared images to avoid the wrong matches during the feature matching.