Code for Dual-stage gradient-guided diffusion model for image super-resolution

Targeting the image super-resolution task, this method employs a diffusion model framework and integrates the Frequency-domain Visual Guidance (F-Guide) module and the Dual-constraint Diffusion Posterior Sampling (D-DPS) module to address structural distortion and hallucinated textures while effectively balancing the fidelity-perception trade-off.

The code primarily serves model training, inference testing, and experimental reproduction, supporting the verification of experimental settings and results presented in the paper.