A directional shock diffusion approach to single image super-resolution

We present a directional shock diffusion approach to single image super-resolution(SR) that is featured by two special treatments to image primitives including edges, corners and junctions. The first is to incorporate an adaptively weighted shock filter to recover the edge sharpness along with the structure-sensitive diffusion terms, and the second is to perform structure oriented diffusion that aims to preserve smooth boundary contours and sharp corners by using multiple orientation estimation. Experimental results show that proposed approach outperforms recent methods for both noiseless and noisy cases on several commonly used image databases.