Efficient and Robust Detection of Duplicate Videos in a Large Database: A Survey

Soni R. Ragho, C. S. Biradar

Abstract: A duplicate video detection method is based on video signature and Fingerprint. Robust or compact frame based descriptor and color layout descriptor used to extract features from images that are constructed from video to create fingerprints which are encoded by vector quantization. Mapping process, measure the distance to find the similarity between query video and video in video database. Then apply a coarse-to-fine matching scheme to check the duplicate video. Vector quantization based model signature retained after dataset pruning to reduce search time for the nearest duplicate video by using pre-computed distance measures tables and discard the near duplicate video using just the partially computed distance from the video to query video. This survey gives an introduction to feature extraction using CLD and explores the various search algorithms, along with the pruning method.

Keywords: Duplicate detection, video fingerprinting, Color layout descriptor, Non-metric distance, Data pruning, and Vector quantization