Realizing super-resolution with a directional anisotropic correlation model

This paper presents a directional anisotropic correlation model for super resolution reconstruction. A local adaptive correlation model is formed by exploiting local edge direction and anisotropic characteristic. For a local region with presence of edge or texture, a directional anisotropic correlation model is suitable to describe the local correlation while the correlation model tends to be more isotropic for a smoother region. A super resolution algorithm is proposed based on the proposed adaptive correlation model. Experiment shows that with the suggested correlation model the proposed algorithm is able to provide a good reconstruction result.