In this paper, we present a novel self-learning single image super-resolution (SR) method, which restores a high-resolution (HR) image from self-examples extracted from the low-resolution (LR) input ...