摘要
提出一种基于相位相关的车载视频稳像算法。采用对数极坐标变换技术以及相位相关技术检测图像帧间的旋转、尺度变换以及平移运动参数。为克服局部运动对全局运动参数的影响,采取了由局部运动参数求取全局运动参数的策略。在建立运动参数状态空间模型的基础上,采用Kalman滤波技术进行运动滤波,获取补偿参数。实验结果说明该方法能够比较理想地保留摄像机的主动运动,去除高频抖动。
An image stabilization algorithm for mobile video surveillance based on phase correlation was presented. Phase correlation and logarithm polar magnitude spectra representation were utilized to obtain translation, rotation and scale parameters. The global motion vector was decided according to the peak values of sub-image phase correlation surfaces to improve the precision. Correction vector was obtained through Kalman filtering of absolute parameters, and image frames were processed. Experimental results for many video sequences show the good performance to display smooth global movements and to eliminate dithering.
出处
《光电工程》
EI
CAS
CSCD
北大核心
2007年第3期37-41,共5页
Opto-Electronic Engineering
关键词
相位相关
全局运动估计
KALMAN滤波
数字稳像
Phase correlation
Global motion estimation
Kalman filtering
Digital image stabilization
