Single image super-resolution based on trainable feature matching attention network

Qizhou Chen, Qing Shao

Abstract: Highlights•A novel Trainable Feature Matching (TFM) module for super-resolution is proposed.•An lightweight improvement for Non-Local (self-attention) is proposed.•Comprehensive experiments is conducted with BI, BD and DN degradation models.•Our method achieves best super-resolution results on multiple benchmarks.