Review and Implementation of High-Dimensional Local Binary Patterns and Its Application to Face Recognition

Bor-Chun Chen, Chu-Song Chen, Winston Hsu

Abstract

High-dimensional local binary patterns have been proved to be a useful feature for face recognition, which provides near-human performance in a widely used face verification benchmark. In this report, we first review the technical aspect of this promising feature, and then we provide our implementation details of the feature. Finally, we show some experimental results using this feature on two public datasets, LFW and CACD.

Publication

Bor-Chun Chen, Chu-Song Chen, Winston H. Hsu. Review and Implementation of High-Dimensional Local Binary Patterns and Its Application to Face Recognition, Technical Report TR-IIS-14-003, Institute of Information Science, Academia Sinica, 2014. [Pdf] [Bibtex]

@techreport{chen2014review,
author={Bor-Chun Chen and Chu-Song Chen and Winston H. Hsu},
title={Review and Implementation of High-Dimensional Local Binary Patterns and Its Application to Face Recognition},
institution={Institute of Information Science, Academia Sinica},
year={2014},
number={TR-IIS-14-003}
}

Results

Performance of HD-LBP on CACD with different PCA dimensions:

ROC curve of HD-LBP on LFW compared with other state-of-the-art methods under unsupervised protocol:

The AUC of different methods on LFW dataset. The HD-LBP can achieve very competitive results compared to other state-of-the-art methods:

Method	AUC
LARK	78.30%
LHS	81.07%
PAF	94.05%
MRF-MLBP	89.94%
HD-LBP	92.11%