Efficient and Robust Detection of Duplicate Videos in a Database

Abstract

In this paper, the duplicate detection method is to retrieve the best matching model video for a given query video using fingerprint. We have used the Color Layout Descriptor method and Opponent Color Space to extract feature from frame and perform k-means based clustering to generate fingerprints which are further encoded by Vector Quantization. The model-to-query video distance is computed using a new distance measure to find the similarity. To perform efficient search coarse-to-fine matching scheme is used to retrieve best match. We perform experiments on query videos and real time video with an average duration of 60 sec; the duplicate video is detected with high similarity.