一种基于光流检测的视频防抖技术  

A Video Anti Shaking Technology Based on Optical Flow Detection

作　　者：陶文 束伟 TAO Wen;SHU Wei(Yangzhou Hangsheng Technology Co.,Ltd.,Yangzhou,Jiangsu,225009,China)

机构地区：[1]扬州航盛科技有限公司,江苏扬州225009

出　　处：《智能城市应用》2023年第6期74-79,共6页Smart City Application

摘　　要：随着智能手机和移动互联网的发展,视频已经成为一种重要的信息载体,视频拍摄过程中难以避免会产生画面抖动,影响用户体验。视频防抖的方法有很多种:机械、光学、电子、软件。文中从软件后处理的方法出发,基于光流法对原始抖动视频进行仿射变换估计,得到每帧画面的水平、垂直、角度偏移量参数,然后使用多项式拟合方法获得平滑的偏移量参数,最后对原始每帧图像使用平滑偏移量参数进行补偿,从而最终获得画面稳定的视频。文中选择python工具对整个防抖处理方案进行实现,并最终达到预期效果。With the development of smartphones and mobile internet,video has become an important information carrier,and it is difficult to avoid screen shake during video shooting,which affects user experience.There are many methods for video anti shake:mechanical,optical,electronic,and software.Starting from the software post-processing method,this paper estimates the affine transformation of the original dithered video based on the optical flow method to obtain the horizontal,vertical and angular offset parameters of each frame,then uses the polynomial fitting method to obtain the smooth offset parameters,and finally uses the smooth offset parameters to compensate the original image of each frame,so as to finally obtain a stable video.The python tool was selected in the article to implement the entire anti shake processing scheme,and the expected results were ultimately achieved.

关 键 词：光流法 视频防抖 特征点 仿射变换 曲线拟合 PYTHON 

分 类 号：TN911.73[电子电信—通信与信息系统]