Single face hallucination via local neighbor patches

Based on learning neighborhood patches a new single face hallucination method is proposed in this paper. In the proposed method, each input low-resolution (LR) position-patch and all patches in a local window centered at the same position of training images are used to hallucinate a high-resolution (HR) face patch, meanwhile two local similarity measurements between each input LR patch and all local LR and HR neighborhood patches of training images are computed to constrain the hallucination. Additionally, a residue image is estimated for the further improvement of the reconstructed result. Experimental results show that the proposed method can obtain superior or competitive results.