C-Face: Using Compare Face on Face Hallucination for Low-Resolution Face Recognition

Face hallucination is a task of generating high-resolution (HR) face images from low-resolution (LR) inputs, which subfield the general image super-resolution. However, most previous methods only consider visual effect, ignoring how to maintain identity face. In this work, we propose novel model, called C-Face network, can generate HR with high quality while preserving information. A recognition network used extract features in training process. order make reconstructed keep information great extent, metric, i.e., loss, proposed. We also new algorithm deal convergence problem. Moreover, since our work mainly focuses on accuracy output, integrate into process ensures that model be real scenarios. Extensive experiments two large scale datasets demonstrate has best performance compared other state-of-the-art methods.