Department of Mathematics

Linear regression has been widely used for face recognition, but many linear regression-based methods are primarily designed for grayscale images and therefore do not fully exploit color information. In this talk, I will present a paper that extends linear regression to color face recognition by formulating the problem in the quaternion domain.

The paper proposes the Quaternion Linear Regression Classifier (QLRC) under this formulation and shows how the quaternion representation can improve recognition performance compared with treating channels independently. Building on QLRC, the authors further introduce the Quaternion Collaborative Representation Optimized Classifier (QCROC), which integrates quaternion linear regression with a quaternion collaborative representation framework into a unified model. Experimental results reported on benchmark datasets demonstrate the effectiveness of QLRC and QCROC for color face recognition.

References:

1. Cuiming Zou, Kit Ian Kou, Li Dong, Xianwei Zheng, and Yuan Yan Tang, From Grayscale to Color: Quaternion Linear Regression for Color Face Recognition, IEEE Access 7, 154131-154140 (2019).