摘要
针对全局运动视频序列采用灰度投影法检测序列帧间运动矢量,具有图像信息利用充分、计算精度高等优点,文中采用了改进的三步搜索算法有效地提高了运动矢量搜索速度,降低了运算复杂度,同时保持了运动矢量的搜索精度。通过帧间运动矢量补偿将相邻三帧图像的相同背景稳定在同一幅图像的相同位上,利用改进的三帧差分法检测运动目标区域。结合数学形态学的闭运算和填充运算等提取出运动目标。实验结果表明,该方法能够快速高效地从背景变化的视频序列中提取运动目标,具有较强的鲁棒性和准确性。
Adopting the gray projection algorithm to detect the inter-frames motion vectors from video sequence with global motion has many advantages, such as sufficingly using the image＇s information and high accuracy. An improved three-step search algorithm is proposed, the proposed algorithm can effectively accelerate motion vectors search and reduce computational complexity while keeping up high search accuracy. The same background of neighbor three frames are fixed to the same position of the same frame through the global motion compensation. Then the improved three-frame difference can be used to detect the moving object area. Using morphologic close-algorithm and morphologic fillalgorithm to extract the moving objects. Experimental results demonstrate that this method can effectively and quickly extract moving objects from the video sequences with changing background, and has strong robustness and accuracy.
出处
《安徽建筑工业学院学报（自然科学版）》
2009年第5期93-98,共6页
Journal of Anhui Institute of Architecture(Natural Science)
基金
安徽省教育厅重点科研项目(KJ2007A053)
关键词
灰度投影
改进的三步搜索算法
全局运动补偿
改进的三帧差分法
the gray projection
improved three-step search algorithm
global-motion compensation
improved three-frame difference
