J4 ›› 2012, Vol. 30 ›› Issue (5): 487-.

• 论文 • 上一篇    下一篇

抖动视频的电子稳像算法

为解决视频序列中出现的抖动问题|提出了一种基于感兴趣区域的电子稳像算法。该算法首先对采集到的每帧图像建立高斯金字塔并进行下采样|然后结合稳像数学模型|求
解仿射变换参数|最后对每帧画面进行运动补偿|输出稳定的图像序列。仿真实验结果表明|该算法具有较好的稳像效果|对较大的旋转和平移运动|有较强的鲁棒性和准确性。与原始序列相比|稳定图像序列帧间的峰值信噪比平均提高了约10 dB。   

  1. 天津大学 电子信息工程学院|天津 300072
  • 出版日期:2012-09-28 发布日期:2012-11-01
  • 作者简介:邹刚(1987—)|男(土家族)|贵州思南人|天津大学硕士研究生|主要从事数字图像与视频处理研究|(Tel)86-18722089679(E-mail)zougang825@163.com;何凯(1972—)|男|沈阳人|天津大学副教授|硕士生导师|主要从事数字图像处理研究,(Tel)86-15510810452(E-mail)hekai626@163.com。
  • 基金资助:

    国家自然科学基金资助项目(61002030)

Electronic Image Stabilization Algorithm of Jitter Video

An efficient algorithm of electronic image stabilization based on ROI(Region Of Interest) was proposed in order to solve the jitter problem in video sequences.Firstly,the Gaussian pyramid is constructed by the captured images and each frame is down-sampled in the following.And then fine estimation is performed by using the optical flow gradient algorithm.So the parameters of affine transform are obtained.Finally,motion compensation is conducted to each frame and the stabilization sequence is acquired.Experimental results show that the subjective effect of the algorithm is good.The algorithm proposed suits for the motion with translation,rotation and scaling while it has better robustness and precision.Compared with the original images before process,the PSNR(Peak Signal-to-Noise Ratio) of the stabled sequences improved approximate 10 dB.   

  1. School of Electronic Information Engineering,Tianjin University,Tianjin 300072,China
  • Online:2012-09-28 Published:2012-11-01

摘要:

为解决视频序列中出现的抖动问题,提出了一种基于感兴趣区域的电子稳像算法。该算法首先对采集到的每帧图像建立高斯金字塔并进行下采样,然后结合稳像数学模型,求
解仿射变换参数,最后对每帧画面进行运动补偿,输出稳定的图像序列。仿真实验结果表明,该算法具有较好的稳像效果,对较大的旋转和平移运动,有较强的鲁棒性和准确性。与原始序列相比,稳定图像序列帧间的峰值信噪比平均提高了约10 dB。

关键词: 电子稳像, 感兴趣区域, 高斯金字塔, 仿射变换

Abstract:

An efficient algorithm of electronic image stabilization based on ROI(Region Of Interest) was proposed in order to solve the jitter problem in video sequences.Firstly,the Gaussian pyramid is constructed by the captured images and each frame is downsampled in the following.And then fine estimation is performed by using the optical flow gradient algorithm.So the parameters of affine transform are obtained.Finally,motion compensation is conducted to each frame and the stabilization sequence is acquired.Experimental results show that the subjective effect of the algorithm is good.The algorithm proposed suits for the motion with translation,rotation and scaling while it has better robustness and precision.Compared with the original images before process,the PSNR(Peak Signal\|to\|Noise Ratio) of the stabled sequences improved approximate 10 dB.

Key words: electronic image stabilization, region of interest(ROI), Gaussian pyramid, affine translation

中图分类号: 

  • TP391.41