AI RESEARCH

Task-Guided Prompting for Unified Remote Sensing Image Restoration

arXiv CS.CV

ArXi:2604.02742v1 Announce Type: cross Remote sensing image restoration (RSIR) is essential for recovering high-fidelity imagery from degraded observations, enabling accurate downstream analysis. However, most existing methods focus on single degradation types within homogeneous data, restricting their practicality in real-world scenarios where multiple degradations often across diverse spectral bands or sensor modalities, creating a significant operational bottleneck.