UNIGE document Chapitre d'actes
previous document  unige:47960  next document
add to browser collection
Title

Robust perceptual hashing as classification problem: decision-theoretic and practical considerations

Authors
Published in Proceedings of the IEEE 9th Workshop on Multimedia Signal Processing, MMSP 2007. Chania (Greece) - Oct 1-3 - IEEE. 2007, p. 345-348
Abstract In this paper we consider the problem of robust perceptual hashing as composite hypothesis testing. First, we formulate this problem as multiple hypothesis testing under prior ambiguity about source statistics and channel parameters representing a family of restricted geometric attacks. We introduce an efficient universal test that achieves the performance of informed decision rules for the specified class of source and geometric channel models. Finally, we consider the practical hash construction, which compromises computational complexity, robustness to geometrical transformations, lack of priors about source statistics and security requirements. The proposed hash is based on a binary hypothesis testing for randomly or semantically selected blocks or regions in sequences or images. We present the results of experimental validation of the developed concept that justifies the practical efficiency of the elaborated framework.
Keywords computational complexityimage sequences
Identifiers
Full text
Structures
Research groups Computer Vision and Multimedia Laboratory
Multimodal Interaction Group
Stochastic Information Processing Group
Citation
(ISO format)
VOLOSHYNOVSKYY, Svyatoslav et al. Robust perceptual hashing as classification problem: decision-theoretic and practical considerations. In: Proceedings of the IEEE 9th Workshop on Multimedia Signal Processing, MMSP 2007. Chania (Greece). [s.l.] : IEEE, 2007. p. 345-348. https://archive-ouverte.unige.ch/unige:47960

171 hits

87 downloads

Update

Deposited on : 2015-03-06

Export document
Format :
Citation style :