UPDF AI

Selective Frequency Network for Image Restoration

Yuning Cui,Y. Tao,5 Autori,Alois Knoll

2023 · DBLP: conf/iclr/0001TBRGC0K23
International Conference on Learning Representations · 186 citazioni

TLDR

A multi-branch and content-aware module is exploited to decompose features into separate frequency subbands dynamically and locally, and then accentuate the useful ones via channel-wise attention weights to handle large-scale degradation blurs.