Log-Gabor Transforms and Score Fusion to Overcome Variations in Appearance for Face Recognition

• Published in: Computer Vision and Graphics Vol. 9972; pp. 353 - 361
• Main Authors: Yildiz, Mithat C̨aǧri, Sharifi, Omid, Eskandari, Maryam
• Format: Book Chapter
• Language: English
• Published: Switzerland Springer International Publishing AG 2016; Springer International Publishing
• Series: Lecture Notes in Computer Science
• Subjects: Class Discriminatory Information; Facial Feature Selection; Feature Transformation Methods; Score Fusion; Sequential Floating Forward Selection (SFFS)
• ISBN: 3319464175; 9783319464176
• ISSN: 0302-9743; 1611-3349
• DOI: 10.1007/978-3-319-46418-3_31

In this paper a new hybrid scheme for overcoming variations in facial images based on the score fusion strategy is considered. The scheme takes into account Log-Gabor transform to extract facial features. The implemented scheme applies Backtracking Search Algorithm (BSA) as a novel feature selection method and Linear Discriminant Analysis (LDA) as a feature transformation method to reduce the number of features and computational cost. Then Weighted Sum Rule (WS) fusion technique is applied to fuse the produced scores for our face recognition system. The robustness of schemes is tested using FERET and ORL database. Experimental results show a significant improvement of proposed scheme over implemented methods in this study.