HIERARCHICAL SELF-ATTENTION-BASED UNIFIED RECONSTRUCTION NETWORK FOR HYPERSPECTRAL IMAGE SUPER-RESOLUTION

Hyperspectral image (HSI) super-resolution aims to improve the spatial resolution of the HSI. Existing methods, particularly those based on convolutional neural networks, often suffer from limited receptive fields, which hinder their ability to capture global features within the image. The lack of global features leads to weakened results for HSI super-resolution. To address these issues, we propose a Unified Reconstruction Network (URC-Net), a deep learning framework that introduces a new self-attention mechanism to autoencoders to effectively fuse HSI and multispectral images. URC-Net improve spatial resolution by capturing global features, and the accuracy of hyperspectral image reconstruction. Extensive experiments across Pavia University datasets validate the effectiveness of our method, showing superior performance compared to existing fusion techniques.