Fast camera fingerprint search algorithm for source camera identification

To determine the source camera of a query image, the fingerprint from the query image needs to be compared with the fingerprints in the reference fingerprint database. Traditionally, the query fingerprint is compared with these reference fingerprints one by one in sequence. For a large database, however, such a brute-force search is inefficient and time-consuming. How to accurately locate the correct fingerprint in the reference fingerprint database is thus becoming a crucial problem for commercial applications of source camera identification. So far there have been few studies in literature addressing this problem. In this work, we propose a new solution to fast fingerprint search. We first store the information of the reference fingerprint digests in the separate-chaining hash table, and then introduce a new rule to select the candidate reference fingerprint digests before performing the correlation. The selection rule is incarnated with the search priority vector. Experimental results have shown that the proposed algorithm outperforms current algorithms.