PolyU Dataset

PolyU Dataset 是一个包含真实世界噪声图像的大型数据集，其中包含合理获得的相应“ground truth”图像。基本思想是多次（例如 500 次）捕获相同且未更改的场景并计算它们的平均图像，这可以粗略地作为真实世界嘈杂图像的“地面真实”图像。该策略的合理性在于，对于每个像素，噪声随机产生大于或小于0。多次采样同一像素并计算平均值将逼近真实像素值并显着减轻噪声。

PolyU Dataset is a large-scale dataset comprising real-world noisy images, paired with appropriately acquired corresponding "ground truth" images. The core idea is to capture the identical, unaltered scene multiple times (e.g., 500 times) and calculate the average of these captured images, which can serve as a rough "ground truth" for real-world noisy images. The rationale behind this strategy is that noise for each pixel randomly fluctuates above or below 0. Sampling the same pixel multiple times and averaging its values will approximate the true pixel value and significantly mitigate noise.

搜集汇总

背景与挑战

背景概述

PolyU Dataset是一个用于真实世界图像去噪的大型数据集，包含带有噪声的图像及其通过多次拍摄平均得到的'ground truth'参考图像。该数据集由香港理工大学和香港中文大学深圳分校于2018年发布，旨在为图像去噪算法提供基准测试资源。