Face recognition of face images with hidden parts using gabor wavelets and PCA

Face Recognition is one of the biometric methods that has recently gained significant attention at the level of research and the scientific community. However, in some special cases, face recognition methods can be sensitive to illumination, facial expressions, aging, face orientation, pose variation and hidden parts of face images, which make recognition very difficult. In this paper, we are interested in the hidden parts of a face image specifically those hidden by facial hair and/or a hair style. We first tested the Eigenfaces method for these modified images; but this approach failed to recognize them. So we have proposed an algorithm that combines Gabor magnitude and phase and PCA. To evaluate the efficiency of our algorithm, we used a variety of face images of FEI Brazilian database and hid the hairstyle and facial hair (barbs) to recognize them from the database images. The obtained results show that the proposed method (Gabor filter and PCA) attained high efficiency in the recognition for this type of problem.