摘要
针对电子稳像运动估计过程中,特征匹配方法存在匹配误差以及3D运动估计方法计算复杂等问题,提出一种基于MEMS(Micro-electro-mechanical system)陀螺仪的电子稳像算法。首先,利用MEMS陀螺仪估计相机旋转抖动量,利用尺度不变特征变换方法SIFT(Scale-invariant feature transform)估计相机平移抖动量;其次,利用滤波技术对平移及旋转运动矢量平滑滤波处理;然后,对视频序列的高频率旋转及平移抖动量进行运动补偿;最后,利用图像拼接技术去除运动补偿后的图像边缘"黑边"。实验结果表明,该算法对于相邻图像帧的抖动补偿准确性较高,处理效果较好,减少了系统的计算复杂度。
To the motion estimation process on the video stabilization, the matching error of feature matching method and calculation complexity of 3D motion estimation method are discussed, a kind of video stabilization algorithm based on MEMS （Micro-electro-mechanical system） gyroscope was proposed. Firstly, the motion es- timation of camera rotation is achieved by using MEMS gyroscope, and the motion estimation of camera transla- tion is achieved by using SIFT （Scale-invariant feature transform）method; next, the filter technology is used to smooth and filter the translation and rotation vector; then, make a camera motion compensation with the vid- eo sequences of the high frequency jitter of rotation and translation ; finally, using image mosaic technology to remove the edge black region of the image after the motion compensation. The experimental resuhs show that the algorithm for the adjacent image frames of jitter compensation accuracy is higher, the effect is good, and reduce the processing time of the system.
作者
赵赛
康宝生
王力
ZHAO Sai, KANG Baosheng, WANG Li(School of hfformation Science ＆ Technology, Northwest University, Xi＇an 710127, Chin)
出处
《西北大学学报（自然科学版）》
CAS
CSCD
北大核心
2018年第3期355-362,共8页
Journal of Northwest University（Natural Science Edition）
基金
国家自然科学基金资助项目(61272286)
