Dual-branch and triple-attention network for pan-sharpening

Pan-sharpening is a technique used to generate high-resolution multi-spectral (HRMS) images by merging high-resolution panchromatic (PAN) images with low-resolution multi-spectral (LRMS) images. Many existing methods face challenges in effectively balancing the trade-off between spectral and spatial information, leading to spectral and spatial structural distortion. In order to effectively tackle these issues, we propose a dual-branch and triple attention (DBTA) network. The proposed DBTA network consists of two essential modules: the Channel-spatial Attention (CSA) module and the Spectral Attention (SPA) module. The CSA module effectively captures the spatial structural information of the images by jointly using spatial and channel attention units. Meanwhile, the SPA module improves the expressive capacity of spectral information by dynamically adjusting channel weights. These two modules work in synergy to achieve comprehensive extraction and fusion of spectral and spatial information, thus resulting in more accurate and clearer reconstructed images. Extensive experiments have been conducted on various satellite datasets to evaluate the performance of the proposed DBTA method outperforms the state-of-the-art competitors in both qualitative and quantitative evaluations.