An adaptive image fusion method for Sentinel-2 images and high-resolution images with long-time intervals

• DOI: 10.1016/j.jag.2023.103381
• Publication Date: 2023-06-07T21:17:02Z
• Authors: Runmin Dong; Lixian Zhang; Weijia Li; Shuai Yuan; Lin Gan; Juepeng Zheng; Haohuan Fu; Lichao Mou; Xiao Xiang Zhu
• Keywords: RGB color model; Feature (linguistics); Land Cover; Spectral bands

Sentinel-2 imagery has garnered significant attention in many earth system studies due to free access and high revisit frequency. Since its spatial resolution is insufficient for applications, e.g., fine-grained land cover mapping, some employ fusion technique that combines high-resolution RGB images with multispectral improve the of latter. However, there are two issues existing image methods. First, these methods usually assume time intervals between short (within several days), which a strong assumption large-scale real-world applications. Second, spectral discrepancy could induce aberrations upon fusion. To alleviate issues, we propose an adaptive approach named S2IFNet, adaptively fusing long-time (from months years) inconsistency, thereby increasing band imagery. Building on top feature extraction modules, compensation module change-aware reconstruction module. The former alleviates possible degradation attributes resulting from latter integrates semantic texture information avoid adding fake textures caused by changes over time. experiments demonstrate S2IFNet surpasses reference-based super-resolution synthetic real datasets, yielding results clearer more reliable.