Coupled Diffusion Posterior Sampling for Unsupervised Hyperspectral and Multispectral Images Fusion.

Area of Science:

• Remote Sensing
• Image Fusion
• Computer Vision

Background:

• Hyperspectral image (HSI) and multispectral image (MSI) fusion is crucial for remote sensing applications.
• Existing deep learning methods often require extensive high-resolution HSI data for supervised training, which is practically scarce.

Purpose of the Study:

• To develop an unsupervised method for HSI and MSI fusion that does not require high-resolution HSI training data.
• To leverage the spectral information from LR-HSI and spatial information from HR-MSI for improved fusion.

Main Methods:

• Proposed a coupled diffusion posterior sampling (CDPS) method for unsupervised HSI and MSI fusion.
• Developed an unsupervised strategy to learn diffusion priors directly from the input LR-HSI and HR-MSI pair.
• Utilized the observed LR-HSI and HR-MSI as fidelity terms within the diffusion posterior sampling framework.

Main Results:

• The CDPS method demonstrated superior performance compared to state-of-the-art unsupervised HSI and MSI fusion techniques.
• The proposed method employs smaller, simpler networks that are easier to train and do not require external datasets.

Conclusions:

• The CDPS method offers an effective unsupervised solution for HSI and MSI fusion, overcoming the data limitations of supervised approaches.
• This technique enables the generation of high-resolution HSI (HR-HSI) without the need for extensive training datasets, making it more practical for real-world applications.