摘要公共安全领域视频图像可能由于环境或自身设备等原因，视频图像会出现抖动，本项目主要针对视频抖动的低质量现象进行稳像处理。本文主要研究视频稳定化的关键技术，研究内容为快速鲁棒光流法，基于光流与多尺度处理的视频稳定化算法及其软件实现。 在综述国内外视频稳定化的相关现状的基础上，总结视频稳像的关键技术，包含两个关键部分：全局运动估计和运动补偿。 （1）在全局运动估计阶段研究了寻找特征点的方法，采用金字塔 Lucas-Kanade 稀疏光流算法进行运动估计，计算出帧间的全局运动矢量。 （2）在运动补偿阶段主要使用运动向量累加法将全局运动向量分离，分离出摄像机正常扫描运动和随机抖动，使用运动补偿向量校正当前帧。实验结果表明这种方法能有效消除视频抖动。 本文开发了基于OpenCV 的数字稳像系统，重点研究了数字稳像关键技术的实现。实验表明，对不同场景的视频，本文所研究的数字稳像系统能够有效消除视频晃动。27579
毕业论文关键词：视频稳像，运动估计，运动补偿，光流法  
Title  Public  Security  Image and  Video  Quality  Enhancement and  Motion Analysis: the Key Technology of Video Stabilization
Abstract Due to the bad environment and the reason of the equipment, images and videos acquired from public security areas  could  appear jitter.  The  project mainly  handles the low quality of the video shaking. The  paper mainly studies the  core  technology of video  stabilization  and  the content  researches the  fast  and  robust optical flow method, which is based on optical flow and multi-scale processing of video stabilization algorithms, and finally software is implemented. In this paper reviewed the related situation of video stabilization at home and abroad, summarizing the key technology of video image, consists of two key parts: global motion estimation and motion compensation.  (1) In the global motion estimation stage find the method of feature points is studied, the pyramid Lucas-Kanade sparse optical flow algorithm for motion estimation, and calculate the global motion vector between the frames. (2) In the Motion compensation phase mainly using the motion vector addition to separate the global motion vector, isolated the normal scan motion of camera and random jitter, using  the  motion compensation vector to correct the current frame. The experimental results show that this method can effectively eliminate the video shaking. This paper developed digital image stabilization system based on OpenCV, focus on the digital image stabilization. The experiment showed that the instability of the video for different scenarios, this paper studies the digital image system can effectively eliminate the video shaking.   Keywords:Video Stabilization,Motion Estimation,Motion Compensation,Optical Flow
目次
1  绪论    ·   1
1.1  研究的目的与意义    1
1.2  视频稳像国内外研究现状    2
1.3  研究内容    2
1.4  论文结构    3
2   视频稳像算法理论      4
2.1  视频稳像算法的基本原理    4
2.2  参数化运动变换数学模型    5
2.3  全局运动估计    5
2.3.1  寻找特征点   ·   6
2.3.2  Lucas-Kanade 稀疏光流法    ·   7
2.3.3  金字塔 Lucas-Kanade光流法  ·   8
2.4  运动补偿    10
2.4.1  求解单应性矩阵   ·  11 公共安全图像与视频一键式质量增强与运动分析视频稳定化关键技术:http://www.751com.cn/jisuanji/lunwen_22094.html