Face Hallucination and Recognition
Face recognition is effective when input probe is of high resolution. In public video surveillance captured face
images are of low resolution. The general methods of face recognition consists steps; super resolving the LR input image
followed by face recognition. Here, we present a framework based on SVD in which we perform face hallucination and
recognition simultaneously. Accuracy of both hallucination and recognition is improved as we perform both of them
simultaneously. We represent each face image by using SVD. For each LR input face image possible images with same
singular values are selected from database. These images are in LR-HR pairs. For reconstructing the HR face images certain
mapping models are learned from selected LR-HR pairs, which are used to interpolate the two matrices in the SVD
representation. This will give more reliable & effective estimation of HR face images.
Index Terms— Face hallucination ; LR Face recognition; Mapping model; S VD.