摘要
全局运动估计是电子稳像的关键技术之一。针对视频大抖动情况下的传统估计方法慢和不准确的问题,提出了一种灰度投影和块匹配相结合的平移运动估计,与基于图像极坐标变换的旋转缩放运动估计方法结合,形成了完整的全局运动估计新方法。使用卡尔曼滤波和视频图像重构,输出新视频。通过对大量视频处理,结果表明该方案提高了全局运动估计方法的运算速度和估计精度,产生了稳定的视频输出。
The global motion estimation is a key technology in video stabilization. Aiming at the drawbacks of traditional estimation method,such as low speed and inaccuracy in the situation of large video jitter,this paper proposes a translational motion estimation method by combining gray projection with block matching algorithm. This method combines with a rotation and zoom motion estimation method based on log-polar reference frame to obtain a new global motion estimation method. The Kalman filter and video image reconstruction is used to produce a new video. The results of many video processing experiments show that the method can improve the computational speed and estimation accuracy of global motion estimation method and output stable video.
出处
《无线电工程》
2016年第2期19-22,共4页
Radio Engineering
基金
国家部委基金资助项目
关键词
视频稳像
全局运动估计
灰度投影法
块匹配法
对数极坐标
video stabilization
global motion estimation
gray projection algorithm
block matching algorithm
log-polar reference frame
