Abstract: Highlights•We propose a fast transformer-based algorithm for light field super-resolution to improve spatial and angular resolution simultaneously.•To extract non-local super-resolution information from different dimensions, we propose a spatial–angular–epipolar swin transformer block.•To overcome the common problem of substantial computational cost and GPU memory consumption in transformer-based methods for 4D light field, we introduce the Unet Swin transformer framework into light field super-resolution.
Loading