A Deep Learning-Based Blind Full Resolution Assessment Method for Pansharpened Images

Pansharpening is the process of integrating multispectral (MS) remote sensing images with panchromatic (PAN) images to generate pansharpened images. The pansharpened image quality assessment(PIQA) is crucial for ensuring the effectiveness of the resulting images. Since the emergence of this research area, it has become a significant focus within the pansharpening field, attracting considerable attention from researchers. However, existing assessment methods often rely on reduced resolution (RR) to obtain reference images, which can lead to a loss of detail. Additionally, due to the lack of available datasets, some researchers adopt full resolution (FR) assessment methods that depend on subjective standards derived from human observation, complicating practical implementation. To address these challenges, we propose a novel no-reference fullresolution image quality assessment (NR-FR IQA) method that does not rely on subjective evaluation standards. Our approach includes a multispectral image feature extraction module to capture spectral features and a panchromatic image feature extraction module to capture spatial features. The features extracted by these two modules will be utilized in the overall assessment model. Experimental results demonstrate that our method performs exceptionally well across various satellite datasets and different fusion methods.