基于改进SIFT算法对录制视频的质量评估

Quality evaluation of recorded video based on Improved SIFT algorithm

文章通过选取30个涵盖人物、风景、设备等不同场景下的短视频,在电视卖场选取9个不同电视机品牌并采用Iphone、VIVO、HUAWEI三种不同拍摄设备对30个短视频进行录制;针对录制视频时出现的位移、形变等问题,采用结合边缘检测及改进SIFT算法的录制视频处理办法,首先使用边缘检测算法寻找播放设备框界,通过逐帧操作实现录制视频空间一致性;然后利用图像特征匹配算法中的尺度不变特征变换匹配视频开始和结束画面,去除时域上产生的冗余;利用录制视频的边缘线性,采用映射边缘检测算法,同时针对视频处理耗时过长的问题,使用曼哈顿距离计算参考图与待匹配图的相似度降低算法复杂度。最后将处理后的录制视频去和原始视频得到空间和时间上的对齐,然后采用SSIM进行质量评估。

In this paper, 30 short videos covering different scenes such as people, scenery and equipment are selected, and nine different TV brands are selected in the TV store, and 30 short videos are recorded by using three different shooting devices:iPhone, vivo and Huawei;aiming at the problems of displacement and deformation in video recording, the video recording processing combining edge detection and Improved SIFT algorithm is adopted Methods: firstly, the edge detection algorithm is used to find the frame boundary of the playback device, and the spatial consistency of the recorded video is realized by frame by frame operation;secondly, the image feature matching algorithm scale invariant feature transformation is used to match the video start and end images to remove the redundancy in the time domain;secondly, the edge linearity of the recorded video is used, and the mapping edge detection algorithm is used, and the video processing time is also considered For the problem of too long, the Manhattan distance is used to calculate the similarity between the reference image and the matching image to reduce the complexity of the algorithm. Finally, the processed recorded video is aligned with the original video in space and time, and then SSIM is used to evaluate the quality.