|
|
|
 |
Kivanc Mihcak
Microsoft Research
Monday, Oct. 31, 11:00am
LC 102, Brooklyn Campus, Polytechnic University
Abstract
We consider the multimedia hashing problem for images:
Given multimedia content as an input, the hash function should produce a
short bit string (termed hash value) such that the hash values are
similar (if not the same) for "perceptually similar" content and
significantly different for "perceptually different" content.
Furthermore, for anti-piracy applications, the hash functions should
have favorable security properties. I will begin the talk with the
definition of the problem of robust signal hashing and some application
scenarios. Then, I will present three image hashing algorithms together
with some experimental results. The first and second ones use
reduced-rank matrix decompositions on randomly chosen image regions; the
first one is based on SVD (singular value decomposition) and the second
one is based on NMF (non-negative matrix factorization). The third
algorithm produces hash values of images via iterative filtering
techniques to collect geometrically significant components of images.
Bio
Kivanc Mihcak was born in Turkey. He received the B.S. degree in
Electrical
Engineering from Bilkent University, Ankara, Turkey, in 1996
(valedictorian). He received full-time scholarship from Bilkent
University during his undergraduate studies. Between 1996-2002, he has
been in the graduate program of the Electrical and Computer Engineering
Department at the University of Illinois, Urbana-Champaign in the Image
Formation and Processing Group, at the Beckman Institute under the
supervision of Pierre Moulin and Kannan Ramchandran. He received the
M.S. and Ph.D. degrees in 1999 and 2002 respectively. Since May 2002, he
is with Microsoft Research, Cryptography & Anti-Piracy Group in Redmond,
Washington.
For further information please contact Nasir Memon (memon at poly.edu)