How to Recognize Facial Images with Spectacles

Two facial images captured from the same person, one with spectacles and another without spectacles, have obviously different appearances. The strong reflection light on spectacles surfaces usually also forms one or several facula regions on facial images. Our proposed method can reconstruct a face image without spectacles frame and facula regions from one with spectacles frame and facula regions. The spectacles frame and facula regions in a facial image can be achieved by means of some simple image processing methods. And principal component analysis (PCA) is then used to obtain a partial reconstruction image while the spectacles frame and facula regions in the facial image are not used. Gabor feature of partially reconstructed image without spectacles frame and facula regions is used for face recognition by means of PCA. Experimental results demonstrate that our method can achieve a higher recognition rate based on the FERET face database