A Physics-based Noise Formation Model for Extreme Low-light Raw Denoising论文学习理解

A Physics-based Noise Formation Model for Extreme Low-light Raw Denoising论文学习

Author: Kaixuan Wei
Link: https://openaccess.thecvf.com/content_CVPR_2020/papers/Wei_A_Physics-Based_Noise_Formation_Model_for_Extreme_Low-Light_Raw_Denoising_CVPR_2020_paper.pdf
Select: ⭐️⭐️⭐️⭐️⭐️
Status: In progress
Type: Academic Journal
code: https://github.com/Vandermode/ELD
备注: 暗光噪声建模
主要将噪声分为四个部分：光子散粒噪声、读取噪声、行噪声和量化噪声。
噪声求取：散粒噪声系统增益；读取噪声形状和尺度参数；行噪声尺度参数。
建模噪声参数的联合分布来求取噪声。
下一篇结合代码看看

Abstract—Enhancing the visibility in extreme low-light environments is a challenging task. Under nearly lightless condition, existing image denoising methods could easily break down due to significantly low SNR. In this paper, we systematically study the noise statistics in the imaging pipeline of CMOS photosensors, and formulate a comp