Face super-resolution reconstruction and recognition from low-resolution image sequences

Normal face recognition algorithms could result in very low recognition rates when the resolution of input faces is very low. In this paper, a new hybrid face recognition algorithm was proposed to tackle the problem of face recognition from low-resolution image sequences. The proposed approach first constructs a high-resolution face image by fully utilizing the sub-pixel information provided in the low-resolution sequences, and then performs Gabor feature based recognition. Experimental results showed that, with respect to applying face recognition to low-resolution face directly, the proposed approach improves recognition rate significantly.