基于L1范数优化路径的视频稳像算法

目的 为解决外包装行业对产品外观进行检测时,采集视频图像存在抖动失真的问题.方法 提出一种基于L1范数优化路径的视频稳像算法,采用SURF算法和FREAK算法对视频序列帧中的特征点进行检测和描述;然后,使用KNN算法和RANSAC算法对相邻帧间的特征点进行匹配和筛选;最后,通过L1范数优化算法对序列帧进行校正和去黑边处理,得到稳像视频.结果 在处理前景无运动和前景有运动的2类视频时,稳像前后视频的平均PSNR值分别提升了5.094 dB和4.273 dB,并且相对于常用的特征匹配算法,文中算法能显著降低相机路径的主动平滑因子.结论 文中算法能够有效地解决视频抖动失真的问题,提高抖动视频的质量,具有一定的鲁棒性.

Video Image Stabilization Algorithm Based on L1 Optimization Path

To solve the problem of jitter and distortion in the collected video images when the outer packaging industry inspects the appearance of the product. This paper proposes a video image stabilization algorithm based on the L1 norm optimization path. The SURF algorithm and the FREAK algorithm are used to detect and describe the feature points in the method frame of the video sequence; then, the KNN algorithm and the RANSAC algorithm are used to compare adjacent feature points between the frames are matched and screened; finally, the sequence frames are corrected and blacked out by the L1 norm optimization algorithm to obtain a stable image. Experiments show that in the two types of videos without foreground motion and with foreground motion, the average PSNR value of the video before and after image stabilization is increased by 5.094 dB and 4.273 dB respectively. Moreover, compared with the common feature matching algorithms, the proposed algorithm can significantly reduce the active smoothing factor of the camera path. The algorithm in this paper can effectively solve the problem of video jitter and distortion, improve the quality of jitter video, and has certain robustness.