In this letter, the problem of feature quantization in robust hashing is studied from the perspective of approximate nearest neighbor (ANN). We model the features of perceptually identical media as ANNs in the feature set and show that ANN indexing can well meet the robustness and discrimination requirements of feature quantization. A feature quantization algorithm is then developed by exploiting the random-projection based ANN indexing. For performance study, the distortion tolerance and randomness of the quantizer are analytically derived. Experimental results demonstrate that the proposed work is superior to state-of-the-art quantizers, and its random nature can provide robust hashing with security against hash forgery.
The copyright of the original papers published on this site belongs to IEICE. Unauthorized use of the original or translated papers is prohibited. See IEICE Provisions on Copyright for details.
Copy
Yue nan LI, Hao LUO, "Approximate Nearest Neighbor Based Feature Quantization Algorithm for Robust Hashing" in IEICE TRANSACTIONS on Information,
vol. E95-D, no. 12, pp. 3109-3112, December 2012, doi: 10.1587/transinf.E95.D.3109.
Abstract: In this letter, the problem of feature quantization in robust hashing is studied from the perspective of approximate nearest neighbor (ANN). We model the features of perceptually identical media as ANNs in the feature set and show that ANN indexing can well meet the robustness and discrimination requirements of feature quantization. A feature quantization algorithm is then developed by exploiting the random-projection based ANN indexing. For performance study, the distortion tolerance and randomness of the quantizer are analytically derived. Experimental results demonstrate that the proposed work is superior to state-of-the-art quantizers, and its random nature can provide robust hashing with security against hash forgery.
URL: https://global.ieice.org/en_transactions/information/10.1587/transinf.E95.D.3109/_p
Copy
@ARTICLE{e95-d_12_3109,
author={Yue nan LI, Hao LUO, },
journal={IEICE TRANSACTIONS on Information},
title={Approximate Nearest Neighbor Based Feature Quantization Algorithm for Robust Hashing},
year={2012},
volume={E95-D},
number={12},
pages={3109-3112},
abstract={In this letter, the problem of feature quantization in robust hashing is studied from the perspective of approximate nearest neighbor (ANN). We model the features of perceptually identical media as ANNs in the feature set and show that ANN indexing can well meet the robustness and discrimination requirements of feature quantization. A feature quantization algorithm is then developed by exploiting the random-projection based ANN indexing. For performance study, the distortion tolerance and randomness of the quantizer are analytically derived. Experimental results demonstrate that the proposed work is superior to state-of-the-art quantizers, and its random nature can provide robust hashing with security against hash forgery.},
keywords={},
doi={10.1587/transinf.E95.D.3109},
ISSN={1745-1361},
month={December},}
Copy
TY - JOUR
TI - Approximate Nearest Neighbor Based Feature Quantization Algorithm for Robust Hashing
T2 - IEICE TRANSACTIONS on Information
SP - 3109
EP - 3112
AU - Yue nan LI
AU - Hao LUO
PY - 2012
DO - 10.1587/transinf.E95.D.3109
JO - IEICE TRANSACTIONS on Information
SN - 1745-1361
VL - E95-D
IS - 12
JA - IEICE TRANSACTIONS on Information
Y1 - December 2012
AB - In this letter, the problem of feature quantization in robust hashing is studied from the perspective of approximate nearest neighbor (ANN). We model the features of perceptually identical media as ANNs in the feature set and show that ANN indexing can well meet the robustness and discrimination requirements of feature quantization. A feature quantization algorithm is then developed by exploiting the random-projection based ANN indexing. For performance study, the distortion tolerance and randomness of the quantizer are analytically derived. Experimental results demonstrate that the proposed work is superior to state-of-the-art quantizers, and its random nature can provide robust hashing with security against hash forgery.
ER -