— This paper presents an extension of the framework proposed by Choi et al. that uses color local texture features (CLTF) for face recognition. The original framework combines multiple CLTFs, each of which corresponds to the associated color channel, with no weights at feature-level. In the proposed extension, the combination is performed with weights that are calculated based on the contribution of each channel to recognition performance. After the combination, PCA is used to select in the combining result the most important components which are used as the unique feature vector representing a face image. Comparative experiments have been conducted to evaluate the customized framework for FR on three public face databases, i.e., CMU-PIE, Color FERET and Postech Face's 01. Experimental results show that the customized framework yields better recognition rates than the framework by Choi et al.
— Color face recognition (FR), color local texture features, color spaces, feature combination, local binary pattern (LBP).
Thanh-Dung Dang is with the Faculty of Information Technology HCMC University of Technical Education Ho Chi Minh City, Vietnam (email: firstname.lastname@example.org).
Xuan-Thu Tuong Thi is with Faculty of Information Technology HCMC University of Foreign Languages – Information Technology Ho Chi Minh City, Vietnam (e-mail: email@example.com).
Cite: Thanh-Dung Dang and Xuan-Thu Tuong Thi, " Weighted Feature-Level Fusion of Color Local Texture Features for Face Recognition," International Journal of Information and Electronics Engineering vol. 5, no. 5, pp. 346-351, 2015.