The Sun-Hays 80 Dataset is a dataset used for super-resolution image research. It uses global scene descriptions to compare and search for related scenes in image databases. These scenes provide ideal example textures to constrain image sampling problems. Compared with internal image statistics for super-resolution tasks, explicit scene matching statistics are more predictive.

The patch-based texture transfer technique produces hallucinated texture details, and then the publisher's super-resolution images are compared with other methods to draw conclusions.

This dataset was released by Brown University in 2012.

Related paper: "Super-resolution from Internet-scale Scene Matching"