Makeup-Invariant Face Recognition using combined Gabor Filter Bank and Histogram of Oriented Gradients

Abstract

Facial makeup is a global problem from the perspective of recognition and security. In this paper, a hybrid feature extraction method is proposed for makeup-invariant face identification and verification. The Gabor Filter Bank (GFB) and Histogram of Oriented Gradients (HOG) were applied to face images from the Virtual Makeup (VMU) database for feature extraction. The final feature vectors were generated through the combination of GFB and HOG features and classified using the City Block Distance (CBD), Euclidean Distance (EUC) and Cosine Similarity Metric (CSM). Performance evaluation of the CBD, EUC and CSM classifiers produced identification and verification rates of 97.39% and 94.12%, 96.73% and 92.16%, and 94.77% and 89.54% respectively for the VMU database. The CSM has the least recognition rate while the CBD achieved the best recognition rates. The implemented method outclassed several face recognition methods previously developed.