Image super-resolution based on guided filter and sparse representation

This article mainly introduces the single image super-resolution (SR) problem based on guided filter and sparse representation. In fact, image super-resolution is highly ill-posed problem, so we needed to regularize it as prior knowledge. The result is to renew a high-resolution image from its down-scale and blurred image. We embark from the recently proposed compressive sensing (CS). We will training high-resolution image and the corresponding low-resolution image patch pairs to generating two over-complete dictionaries D h and D l . In this paper, we exploited guided image filtering as the feature extraction for the low-resolution image patch, instead of the second-order and first-order derivatives. We will showing the results with original images both visual and image PSNR improvements.