SDG-DSB: Spectral Degradation Guided Diffusion Schrödinger Bridge for Hyperspectral Images Super-Resolution

This is the official pytorch code for "SDG-DSB: Spectral Degradation Guided Diffusion Schrödinger Bridge for Hyperspectral Images Super-Resolution", which has been accepted by TGRS2025.

Getting started

Installation

conda env create --file requirements.yaml python=3
conda activate sdgdsb

pre-trained checkpoints

Download the weight file 256x256_diffusion_uncond.pt from Google Drive to the data directory, and download latest.pt to the results/sr4x-bicubic directory.

Running

python sample.py --clip-denoise --use-cddb-deep

Acknowledge

Some of the codes and pre-trained checkpoints are built upon CDDB and I2SB.