--- license: unknown --- # Description This dataset contains a paired real hyperspectral image (HSI) denoising dataset, which consists of shortexposure noisy HSIs and the corresponding long-exposure clean HSIs. The dataset employs a SOC710-VP hyperspectral camera, manufactured by Surface Optics Corporation (SOC), USA. The SOC710-VP hyperspectral camera is equipped with a silicon-based charge-coupled device (CCD) and an integrated scanning system. With standard settings, the SOC710-VP can capture HSI with 696 × 520 pixels in spatial resolution and 256 spectral bands from 376.76 nm to 1075.80 nm at 2.7 nm interval. The dataset contains both indoor and outdoor scenes. Indoor images under incandescent lamp to lighten the scene. The outdoor images are generally captured on sunny days between around 9am and 5pm. 34 bands were selected around 400nm to 700nm in the visible spectral range. The dynamic range of the captured HSI is 12 bit, and the spectral value ranges from 0 to 4095. The dataset contains 62 image pairs, separated as follows into two folders: - `gt`: where the camera settings, such as aperture, focus and exposure time, are adjusted to maximize the quality. - `input50`: short-exposure image, where the exposure time was intentionally decreased by a factor of 50 with respect to the reference. ## Credits Dataset originally collected at this Google drive link: Associated repository: If this dataset is used for research, please ensure to cite this paper: ```bibtex @InProceedings{Zhang_2021_ICCV, author = {Zhang, Tao and Fu, Ying and Li, Cheng}, title = {Hyperspectral Image Denoising With Realistic Data}, booktitle = {Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV)}, month = {October}, year = {2021}, pages = {2248-2257} } ```