为了实时地稳定摄像系统的输出视频,提出了一种基于特征匹配的实时电子稳像算法。将改进 的Kanade2Lucas2Tomasi (KLT) 算法用于特征提取,并提出一种新的基于灰度投影均值的特征匹配算法。为了保 证稳像结果的鲁棒性,还给出了特征的有效性检验方法。-Real-time camera system in order to stabilize the output of video, presents a feature-based matching algorithm for

该算法可以在视频大幅抖动含有运动目标时和全局运动时进行准确稳像-The algorithm can in big video jitter contain moving targets and global movement to stabilize the image accurately