Deep Image Super Resolution via Natural Image Priors.

Single image super-resolution (SR) via deep learning has recently gainedsignificant attention in the literature. Convolutional neural networks (CNNs)are typically learned to represent the mapping between low-resolution (LR) andhigh-resolution (HR) images/patches with the help of training examples. Mostexisting deep networks for SR produce high quality results when training datais abundant. However, their performance degrades sharply when training islimited. We propose to regularize deep structures with prior knowledge aboutthe images so that they can capture more structural information from the samelimited data. In particular, we incorporate in a tractable fashion within theCNN framework, natural image priors which have shown to have much recentsuccess in imaging and vision inverse problems. Experimental results show thatthe proposed deep network with natural image priors is particularly effectivein training starved regimes.