Face recognition using Complete Gabor Filter with Random Forest

Abstract

This paper proposes a hybrid face recognition technique called Complete Gabor Classifier with Random Forest (CGC-RF) in biometrics technologies. CGC-RF uses Gabor Filter and Oriented Gabor Phase Congruency Image (OGPCI) with Random Forest as the learning framework. The Gabor Filter provides the magnitude information of Gabor responses, where the OGPCI contains the phase information of Gabor response. Random Forest is used as the learning framework to classify images based on the features extracted from both Gabor Filter and OGPCI. We tested the proposed technique by assessing the face verification and identification on two face databases namely, the Georgia Tech Face and Faces94. These databases consisted of face image with varied characteristics such as head positions, head orientations, occlusion and light illumination. The results of the assessment suggest the proposed CGC-RF produced high recognition rates of face images on all two databases. It is of our view that GGC-RF outperformed existing face recognition techniques such as PCA, LDA and Gabor-PCA.