欢迎您！东篱公司
退出
-
申报数据库
-
立项数据库
-
成果数据库
-
项目获奖数据库
中文摘要：
针对数字视频帧间平移抖动的稳定问题,介绍一种基于局部求精位平面匹配运动估计和约束卡尔曼滤波运动校正的视频稳定算法。运动估计首先结合了灰阶比特平面匹配和菱形搜索策略得到初步的估计结果,然后在其附近再以最小绝对差（MAD）为测度,搜索更为准确的运动估计结果。这种运动估计方法在保证估计精度的前提下,显著地减少了运动估计需要的计算量。运动校正则考虑到实际稳像系统对校正量可能存在的某些约束,对绝对帧位移曲线采用约束卡尔曼滤波,得到平滑的位移曲线,有效地降低了帧间抖动的幅度,同时保证了校正矢量不超过稳像系统的实际校正能力。仿真实验表明,该算法具有精度高、速度快的特点,尤其适用于实时视频稳定。
英文摘要：
Aiming to stabilize translational jitters between digital video frames,an image stabilization algorithm based on a locally refined gray-coded bit-plane matching motion estimation and constrained Kalman filter motion compensation was presented. Motion estimation firstly combined gray-coded bit- plane matching and diamond search strategy to acquire preliminary estimation results, then catched out further estimation to find out more precise results around these preliminary ones with mean absolute difference as the motion estimation measure. The motion estimation method could greatly reduce computational load while maintaining motion estimation precision.In viewing of some possible constrains imposed by practical image stabilization system, motion compensation adopted constrained Kalman filter to filter absolute displacement curve that effectively reduced the altitude of jitters between video flames and obtained smooth displacement curve, while the compensation vector was not beyond the actual compensation ability of the image stabilization system. Simulation experiments show the algorithm is simple, fast and effective, especially fits for real-time video stabilization.
同期刊论文项目
同项目期刊论文
期刊信息
位置：