| 摘要: |
| 目的 解决检测摄像机在拍摄过程中因外界抖动导致视频扭曲失真等问题。方法 选取视频第1帧作为基准帧，检测视频每帧的特征点，并与第1帧特征点位置进行对比，经过仿射变换消除每一帧的扰动。通过FAST算法生成每一帧特征点，利用BRISK算法得到特征点的二进制特征描述符，比较特征点描述符的汉明距离来匹配相邻两帧间的特征点，利用匹配好的特征点对确定仿射变换模型，最后利用得到的仿射变换矩阵对每一帧进行校正。结果 实验原视频总帧数为23帧，时长为1 s，分辨率为320×240，视频中除第1帧外，其余22帧均得到修正。结论 FAST角点检测算法与BRISK匹配方法结合可以有效解决视频扭曲等问题。 |
| 关键词: 图像处理 视频稳定 运动校正 特征点检测 仿射变换 |
| DOI：10.19554/j.cnki.1001-3563.2018.21.039 |
| 分类号:TP391.4 |
| 基金项目: |
|
| Video Stabilization Algorithm Based on Feature Point Matching |
|
MA Zhong-xue, MU Ping-an, DAI Shu-guang
|
|
School of Optical-Electrical and Computer Engineering, University of Shanghai for Science and Technology, Shanghai 200093, China
|
| Abstract: |
| The paper aims to solve the problem that the detection camera is distorted due to external jitter during the shooting process. The first frame of the video was selected as the reference frame, and the feature points of each frame of the video were detected. Compared with the position of the feature points of the first frame, and the perturbation of each frame was eliminated by affine transformation. The FAST algorithm was used to generate frame feature point. The BRISK algorithm was used to obtain the binary feature descriptor of the feature point. The Hamming distance of the feature point descriptor was compared to match the feature points between the adjacent two frames; and the affine transformation model were determined with matched feature point pairs. The affine transformation model was finally corrected for each frame with the obtained affine transformation matrix. The total number of frames of the original video was 23 frames, the duration was 1 second, and the resolution was 320×240. All frames except the first frame were corrected. The FAST corner detection algorithm combined with the BRISK matching method can effectively solve problems such as video distortion. |
| Key words: image processing video stabilization motion correction point-feature detecting affine transformation |
