SNIC: Synthesized Noisy Images using Calibration Dataset

This dataset supports a forthcoming arXiv preprint. The dataset consists of over 6,000 images from a total of four camera sensors. For every camera / scene, there are clean (low ISO) ground truth images as well as noisy images at a range of ISO values. The noisy images were synthetically generated using calibrated, tuned noise models. The resulting noisy images are provided in both RAW and TIFF format, as are the ground truth images. The code to generate these images can be found here: https://github.com/nikbhatt-cu/SNIC.git

本数据集为一篇即将发布的arXiv预印本提供研究支持。该数据集包含来自4款相机传感器的6000余张图像。针对每款相机与每个拍摄场景，均配有低ISO的纯净真值图像（ground truth），以及一系列不同ISO取值下的带噪图像。带噪图像通过经校准与调优的噪声模型合成生成。生成的带噪图像与真值图像均提供RAW与TIFF两种格式。用于生成该数据集的代码可通过以下链接获取：https://github.com/nikbhatt-cu/SNIC.git