基于抖动帧检测和低通滤波的数字视频稳像算法  

Digital video stabilization based on jittery frame checking and low-pass filtering

作　　者：许杰[1] 杨硕[1] 王明辉[1] 

机构地区：[1]四川大学计算机学院,成都610065

出　　处：《四川大学学报（自然科学版）》2013年第6期1235-1240,共6页Journal of Sichuan University(Natural Science Edition)

基　　金：国家自然科学基金(61071162)

摘　　要：近年来,数字视频稳像方法是提高视频质量的重要研究内容之一.针对高斯滤波使用固定高斯核进行运动平滑,去除抖动能力有限的问题,提出根据视频帧抖动程度使用不同低通滤波进行运动平滑的方法.首先,计算当前帧与其前后固定半径范围内相邻帧之间的位移量,并计算其位移方差,该方差可以有效地描述视频帧的抖动程度;其次,根据抖动程度使用自适应滤波或均值滤波进行运动平滑.实验证明,该方法比现有的方法具有更好的稳像效果.In recent year, digital video stabilization is one of the important research fields of video quality improvement. To solve the inefficient of Gaussian filtering that perform motion smoothing by a constant Gaussian kernel, a novel motion smoothing method was proposed. Firstly, the displacements between current frame and its neighboring frames are calculated, the variance of the displacements is also calcu- lated, which can measure the jitter of frame effectively. Then, smoothed motion can be obtained by an adaptive filter or a mean filter with respect to the variance. The experimental results showed that pro- posed method is superior to the existing method.

关 键 词：数字视频稳像 运动平滑 自适应滤波 均值滤波 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]