Polarization Denoising and Demosaicking Dataset
收藏资源简介:
本数据集包含40个真实场景,并提供了三种不同噪声水平下的数据。每个场景都包含了一对数据:一个是有噪声的马赛克输入图像,另一个是与之对应的噪声-free完整图像。该数据集旨在为极化图像的去噪和去马赛克任务提供一个新的评估平台,有助于推动该领域的研究。数据集的创建过程包括使用JAI CV-M9GE 3-CCD RGB相机和SIGMAKOKI SPF-50C-32线性偏振滤光片,通过旋转偏振滤光片在0度、45度、90度和135度四个角度下捕获图像,然后通过平均大量捕获的图像来生成噪声-free的参考图像。数据集可用于评估和开发极化图像的去噪和去马赛克算法,特别是在动态场景中的应用。
This dataset includes 40 real-world scenes, with data provided under three distinct noise levels. Each scene contains a pair of data: one is a noisy mosaic input image, and the other is the corresponding noise-free complete image. This dataset aims to provide a novel evaluation platform for denoising and demosaicing tasks of polarization images, facilitating research advancement in this field. The dataset was constructed using a JAI CV-M9GE 3-CCD RGB camera and a SIGMAKOKI SPF-50C-32 linear polarizing filter. Images were captured by rotating the polarizing filter at four angles: 0°, 45°, 90°, and 135°, followed by averaging a large set of captured images to generate the noise-free reference images. This dataset can be utilized to evaluate and develop denoising and demosaicing algorithms for polarization images, particularly for applications in dynamic scenarios.

- 1通过东京科学研究所,日本 · 2025年



