Image super-resolution based on Null Space Pursuit

In this paper, we propose an image super-resolution algorithm using the Null Space Pursuit (NSP) and Local Gaussian Model (LGM). The NSP is a method for signal decomposing, which allow us to separate a signal into a summation of narrow band signals. Firstly, an image is decomposed into narrow band part and its residual, then enlarge the narrow band part with interpolation and specify the LGM as the prior distribution model of the residual part, and formulate the image super-resolution problem as a constrained optimization problem to obtain the enlarged residual part. Finally, we get super-resolution image by combining enlarged narrow band and its residual parts.