"Remote Sensing Destriping Dataset"

"This dataset is specifically designed for the evaluation of destriping methods in multispectral remote sensing imagery, and it comprises both simulated and real-world data. The dataset is provided in two formats, namely .tiff and .mat, to accommodate different experimental requirements. The .tiff images are 8-bit quantized with a dynamic range of [0, 255], which preserves the original imaging characteristics commonly used in visualization and traditional processing pipelines. In contrast, the .mat images are stored in double-precision format and normalized to the range [0, 1], making them more suitable for numerical computation and machine learning-based approaches.The simulated data are generated by cropping clean sub-images from multiple satellite sources, including MODIS, CHRIS, and HY-1E sensors. To ensure the diversity and realism of stripe noise patterns, both periodic and non-periodic stripe noise are artificially introduced into the clean images. This allows for a controlled evaluation of destriping performance under different noise conditions.In addition to simulated data, the dataset also contains real-world stripe-contaminated imagery collected from three distinct remote sensing platforms. Specifically, these include a 400\u00d7400 sub-image from Aqua MODIS Band 30, a 300\u00d7300 sub-image from GF5-AHSI Band 42, and a 400\u00d7400 sub-image from HY-1E-PMRIS Band 17. These real data samples reflect practical imaging degradations and provide a benchmark for assessing the robustness and generalization capability of destriping algorithms in real applications.Overall, the dataset offers a comprehensive testbed that integrates controlled simulation and real observational data, enabling thorough and reliable evaluation of destriping techniques across diverse scenarios."