Hybrid CNN-transformer network with frequency-aware fusion for efficient single image super-resolution
Abstract: Highlights•We proposed FAFB to flexibly extract and fuse high- and low-frequency features for improved representation.•We designed Hybrid-former with adaptive split ratio in FAFB to balance CNN and Transformer strengths efficiently.•Extensive experiments validate the effectiveness of FAFB and Hybrid-former for SISR tasks.
External IDs:dblp:journals/eswa/ZhouZSCWG25
Loading