Efficient video retrieval using shot clustering and alignment

Retrieving similar video clips from large video database requires a fast methods that can attain high precision and recall. This paper adopts clustering to speed up the search in database, and proposes a new distance measure, which aligns two video clips to improve the accuracy of their similarity. Based on the clustering and the distance measure, we propose a new algorithm to construct candidate video clips in video query by example. Experimental results show that the proposed video retrieval method is efficient.