Graphics Reference
In-Depth Information
Infrared Face Recognition Based on DCT and Partial
Least Squares
Zhihua Xie and Guodong Liu
Key Lab of Optic-Electronic and Communication,
Jiangxi Sciences and Technology Normal University, Nanchang, Jiangxi, 330013, China
xie_zhihua68@aliyun.com
Abstract. Infrared face imaging, being light- independent, and not vulnerable to
facial skin expressions and posture, can avoid or limit the drawbacks of face
recognition in visible light. However, to obtain the compact and discriminative
feature extracted from infrared face image is a challenging task. In this essay,
infrared face recognition method using Discrete Cosine Transform (DCT) and
Partial Least Square (PLS) is proposed. Due to strong ability for data de-
correlation and compact energy, DCT is studied to obtain the compact features
in infrared face. To make full use of the discriminative information in DCT
coefficients, the final classifier formulates PLS regression for accurate
classification. The experimental results show that the proposed algorithm
outperforms Principle Component Analysis (PCA) and DCT based infrared face
recognition algorithms.
Keywords: Infrared face recognition, Partial least square, feature extraction,
discrete cosine transform.
1
Introduction
As we know, the resolution of the infrared image is lower than the visible image. This
is to say that the infrared image has little local discriminative information.
For this reason, the compact and discriminative feature extraction from the infrared
face image is a challenging task [4, 8]. In previous research, Discrete Cosine
Transform (DCT) based on the feature extraction method is applied to extract
compact information for face recognition [5]: Zhang et al [7] improved the classical
face features extraction method (Principle Component Analysis (PCA) +Linear
Discriminant Analysis (LDA)) and proposed DCT and LDA based on features
extraction algorithm; Yin et al [6] improved DCT and LDA face recognition method
using Feature Selection (FS) in DCT domain. As for infrared face recognition, Xie et
al [8] applied DCT and FS to find a compact features extraction method. However,
the discriminative performance of DCT features received less attention. The main
idea in this essay is that different DCT coefficients do have different ability to
discriminate various classes. In other words, some coefficients, namely discriminant
coefficients, should have bigger weights than others. Therefore, how to make full use
 
Search WWH ::




Custom Search