• 中国计算机学会会刊
  • 中国科技核心期刊
  • 中文核心期刊

J4 ›› 2006, Vol. 28 ›› Issue (1): 82-85.

• 论文 • 上一篇    下一篇

基于投影特征的快速图像稳定算法

黄金权 王程 王润生   

  • 出版日期:2006-01-01 发布日期:2010-05-20

  • Online:2006-01-01 Published:2010-05-20

摘要:

本文针对摄像机的栽体既有抖动又有正常扫描运动的图像稳定问题提出了一种快速算法。该算法首先把二雏图像数据投影为一维数据,并在投影数据上估计图像序列的偏移量;然后通过卡尔曼滤波从图像序列的偏移量序列中消除抖动带来的影响,从而保持正常的扫描运动;最后依据滤波后的偏移矢量对图像进行运动补偿。实验结果表明,该算法
法在保持与块匹配算法相当的稳像准确度的前提下,大大减少了计算量。

关键词: 图像稳定 运动估计 卡尔曼滤波 运动补偿

Abstract:

To solve the image stabilizing problem due to the carrier of videocams having the dithering vectors and the common motion vectors, we present a fast a lgorithm. First, it convcrts the 2-D data of the image sequences into 1-D data, and it makes use of the 1-D data to estimate the offsets of the image se quences. Second, it keeps the corranon motion vectors and removes the dithering vectors through the Kalman filter. Finally, we deal with the current ima  ge according to its offset vector. Experimental results indicate that it reduces the computational load greatly, while the accuracy does not degrade.

Key words: image stabilizing;motion estimation, Kalman filter, motion compensation