Recently two-dimensional (2D) PalmHash Code, whose original features were extracted by 2D Gabor filter, was proposed as a cancelable palmprint coding scheme for secure palmprint verification. However, vertical correlation inherently found in 2DPalmHash Code is vulnerable to statistical analysis attack. It was empirically attested that the vertical correlation of 2DPalmHash Code was suppressed by transposing Gabor feature matrices that were tuned to vertical orientation. Unfortunately, the mechanism behind why it works remains open. In this paper, we first systematically analyze and reveal the horizontal and vertical correlations of 2DPalmHash Code. We then uncover the secret behind the transposition of Gabor feature matrices for vertical correlation suppression. Finally, the orientation range of Gabor feature matrices suitable for transposition is determined from the correlation analysis and verified by a series of experiments.
Bibliographical noteFunding Information:
This work was partially supported by Basic Science Research Program through the National Research Foundation (NRF) of Korea funded by Ministry of Science, ICT and Future Planning ( 2013006574 ), National Natural Science Foundation of China (61305010, 61262019, 61202112, and 61303199), Research Group Project funded by Deanship of Scientific Research at King Saud University ( RGP-VPP-288 ), China Postdoctoral Science Foundation ( 2013M531554 ), and Doctoral Initiating Foundation of Nanchang Hangkong University ( EA201308058 ).
All Science Journal Classification (ASJC) codes
- Computer Science Applications
- Cognitive Neuroscience
- Artificial Intelligence