IR275K: A Benchmark for Infrared Multi-Frame Super-Resolution Toward Efficient Remote Sensing

Researchers introduced IR275K, a curated infrared video benchmark with 594 sequences and 275,196 frames for multi-frame super-resolution.

The benchmark includes a reproducible x4 evaluation protocol, aiming to standardize comparisons under realistic sensor and platform constraints.

They also evaluated CGMamba, a lightweight model that delivered strong image quality with relatively low compute.

Results suggest that explicit spatial encoding, including 2D rotary position encoding, can improve infrared super-resolution performance.