Near Duplicate Video Detection Algorithm Based on Gray-scale Pyramid and Temporal Features

Near duplicate video detection algorithms based on global feature is sensitive to the brightness and cropping transform of videos. To solve the problem, a near duplicate video detection algorithm based on gray-scale pyramid and temporal feature called GSPOM is proposed. Before extracting the block brightness feature of the video, select the first k frames of the test video and match the gray-scale pyramid feature with the query ones. On the one hand, the dissimilar video will be filtered by matching and the amount of data will be reduced; On the other hand, the editing modes of the test video are estimated by the matching process. The corresponding gray-scale feature of the keyframe in the video library is extracted as the keyframe, and its features are extracted and matched with the feature of the test video. The similarity between them is also calculated. Experiments show that the algorithm has a significant improvement compared with the traditional algorithm based on temporal features.