CGFformer: A New AI Method for Pansharpening Satellite Images
A novel approach known as CGFformer (Cluster-Guidance Frequency Transformer) has been introduced for the process of pansharpening, which merges low-resolution multispectral (LRMS) images with high-resolution panchromatic (PAN) images to create high-resolution multispectral (HRMS) images. Traditional frequency-based pansharpening techniques rely on static filters that struggle with complex frequency distributions, and current denoising methods often fall short in managing different types of noise. CGFformer overcomes these challenges through an adaptive separation module that employs K-means clustering to combine local and non-local data, allowing for enhanced separation of high- and low-frequency elements. This method emphasizes the variability of frequency distribution and the interplay between frequency and spatial components. The findings were shared on arXiv under the identifier 2605.01490.
Key facts
- CGFformer stands for Cluster-Guidance Frequency Transformer
- It is designed for pansharpening, fusing LRMS and PAN images into HRMS images
- Current methods use fixed frequency filters that cannot adapt to complex distributions
- Existing denoising strategies insufficiently exploit frequency components
- CGFformer uses an adaptive separation module with K-means clustering
- It integrates local features and non-local information
- The method enables more precise separation of high- and low-frequency components
- The research was published on arXiv with identifier 2605.01490
Entities
Institutions
- arXiv