Fig. 1From: Single image super resolution based on multi-scale structure and non-local smoothingFlowchart of the proposed SR method. s is the size ratio between the output HR and the input LR; si is the image magnification factor in the ith iteration, and it determines the choice of the global dictionaries \(D^{H}_{{si}}\) and \(D^{L}_{{si}}\); \(D^{0}_{{si}}\) is the local dictionary trained from the image itselfBack to article page