Abstract: Promising research results have emerged in recent years in identifying the source acquisition device of multimedia objects. These efforts have primarily focused on the design of techniques that can identify and extract class properties and individual properties of images and videos. Although the overwhelming amount of multimedia information is one of the most significant challenges faced by source-device identification techniques,current techniques did not considered the computational scalability as a concern. In this paper, we propose a new approach, which enables fast retrieval of media objects that exhibit certain forensically relevant characteristics, most importantly PRNU based imaging sensor fingerprints. Proposed binary search tree (BST) based retrieval scheme offers logarithmic improvements in the efficiency of the conventional techniques without a significant reduction in the performance.

@inproceedings{bayram-sencar-memon-spie10,
  url          = {http://spiedl.aip.org/getabs/servlet/GetabsServlet?prog=normal&id=PSISDG007541000001754109000001&idtype=cvips&gifs=yes&ref=no},
  booktitle    = {SPIE Conference on Media Forensics and Security},
  author       = {Sevin\c{c} Bayram and Husrev T. Sencar and Nasir D. Memon},
  location     = {San Jose, CA},
  year         = {2010},
  title        = {Efficient techniques for sensor fingerprint matching in large image and video databases},
}