摘要
针对位平面匹配算法很难对视频图像中存在的旋转运动进行估计的缺陷,提出了一种改进的电子稳像算法。该算法采用钻石搜索策略的位平面法来估计并补偿视频图像序列间的平移运动,对参考帧图像进行拉普拉斯变换,在靠近图像边缘处选取若干个特征小块与当前帧进行块匹配,进一步计算并补偿其可能存在的旋转运动。通过理论分析和实验仿真表明:在处理存在旋转运动的视频图像时,改进后的算法在满足实时性的同时提高了稳像精度。
Aiming at the bit-plane algorithm＇s defect of hardly estimating the rotational motion among video images,an improved electronic image stabilization is presented.The translational motion between image sequences is calculated and compensated based on bit-plane algorithm with diamond search strategy.Near the side of the image, several blocks with distinguishing characteristic are selected by Laplacian transform,the potential rotational motion is also calculated and compensated by block matching algorithm.Through theoretical analysis and simulation experiments,the improved method satisfies the real-time demand,and has the better accuracy of image stabilization in rotational motion than former.
出处
《计算机工程与应用》
CSCD
北大核心
2011年第28期179-181,共3页
Computer Engineering and Applications
关键词
电子稳像
位平面法
块匹配
旋转运动
electronic image stabilization
bit-plane
block matching
rotational motion
