Raw Natural Image Noise Dataset (RawNIND)
收藏资源简介:
RawNIND是由鲁汶大学创建的原始图像噪声数据集,旨在支持跨传感器、图像处理流程和风格的去噪模型开发。该数据集包含2831对原始图像,涵盖了多种相机传感器和噪声条件。数据集的创建过程包括在较高ISO设置下捕捉噪声图像,并在基础ISO和最佳曝光设置下捕捉对应的干净图像,以确保场景和光照条件的一致性。数据集的应用领域主要集中在图像去噪和压缩,旨在解决传统方法在处理已开发图像时的性能限制和泛化问题,提升图像处理的效率和灵活性。
RawNIND is a raw image noise dataset developed by KU Leuven, designed to support the development of denoising models across various camera sensors, image processing pipelines and styles. This dataset includes 2,831 pairs of raw images, covering a wide range of camera sensors and noise conditions. The dataset was constructed by capturing noisy images at high ISO settings, alongside corresponding clean images acquired at base ISO and optimal exposure settings, to guarantee consistency in scene and lighting conditions. Its primary application areas lie in image denoising and compression, aiming to address the performance limitations and generalization issues of traditional methods when processing developed images, and enhance the efficiency and flexibility of image processing.

- 1Learning Joint Denoising, Demosaicing, and Compression from the Raw Natural Image Noise Dataset鲁汶大学 · 2025年



