Light field images record a series of viewpoints of a scene and thus have many attractive
applications. However, the trade-off between angular and spatial resolution in the imaging
process makes light field image super-resolution necessary. In this paper, we propose a
Content-Aware Spatial-Angular Interaction (dubbed CASAI) network for light field image
super-resolution. The gradient branch of CASAI makes full use of the context information of
the low-resolution gradient map and the multi-level features of the super-resolution branch�…