single image super-resolution (SISR), as a traditional ill-conditioned
inverse problem, has been greatly revitalized by the recent development of
convolutional neural networks (CNN). These CNN-based methods generally map a
low-resolution image to its corresponding high-resolution versi