2021/09/29 by Nitin Balachandran, Jun-Cheng Chen, Balachandran, Nitin +3
Computer Science · Medicine · #Advanced Image Processing Techniques #Aesthetics #Ambiguity #Art #Artificial intelligence #Computer Vision and Pattern Recognition (cs.CV) #Computer science #Computer vision #FOS: Computer and information sciences #Face (sociological concept) #Face detection #Face hallucination #Face recognition and analysis #Facial Nerve Paralysis Treatment and Research #Facial recognition system #Identity (music) #Image (mathematics) #Image and Signal Denoising Methods #Leprosy Research and Treatment #Pattern recognition (psychology) #Process (computing) #Resolution (logic) #cs.CV
paper · pdf · doi:10.48550/arxiv.2109.14690
published in arXiv (Cornell University) (Cornell University)
arxiv created 2021/09/29 · openalex publication_date 2021/09/29 · arxiv updated 2021/10/01 · openalex created_date 2025/10/10 · openalex updated_date 2026/08/08
Face super-resolution is a challenging and highly ill-posed problem since a\nlow-resolution (LR) face image may correspond to multiple high-resolution (HR)\nones during the hallucination process and cause a dramatic identity change for\nthe final super-resolved results. Thus, to address this problem, we propose an\nend-to-end progressive learning framework incorporating facial attributes and\nenforcing additional supervision from multi-scale discriminators. By\nincorporating facial attributes into the learning process and progressively\nresolving the facial image, the mapping between LR and HR images is constrained\nmore, and this significantly helps to reduce the ambiguity and uncertainty in\none-to-many mapping. In addition, we conduct thorough evaluations on the CelebA\ndataset following the settings of previous works (i.e. super-resolving by a\nfactor of 8x from tiny 16x16 face images.), and the results demonstrate that\nthe proposed approach can yield satisfactory face hallucination images\noutperforming other state-of-the-art approaches.\n