从黑暗中学习基于物理的噪声模型,用于低光原始图像去除
IEEE transactions on pattern analysis and machine intelligence
|January 12, 2026
概括
这项研究引入了一种新的方法,用于在低光下消除图像模糊,从黑暗的中学习噪声模型,而不是对联的真实数据. 这种方法提高了合成数据的准确性,以提高现实世界的性能.
科学领域:
- 计算机视觉 计算机视觉
- 图像处理 图像处理
- 机器学习 机器学习
背景情况:
- 目前的低光原始图像去暗化严重依赖于合成数据.
- 现有的噪声建模方法 (基于物理和基于学习) 在准确性和数据依赖性方面存在局限性.
- 有效的噪声建模对于无噪声算法的实际应用至关重要.
研究的目的:
- 开发一种新的策略,通过学习来自黑暗的噪声模型来训练低光消光方法,减少对配对真实数据的依赖.
- 引入一个高效的基于物理学的噪声神经代理 (PNNP) 进行准确的现实世界传感器噪声建模.
- 为了提高合成数据的有效性和实用性,用于低光原始图像无光化.
主要方法:
- 提出了一种从黑暗中学习噪声模型的策略,消除了对真实数据配对的需求.
- 介绍了基于物理的噪音神经代理 (PNNP),将物理先验集成到神经网络中.
- 开发了三个关键技术:物理引导的噪声解 (PND),物理意识的代理模型 (PPM) 和可差分发损失 (DDL).
主要成果:
- PNNP有效地描述了现实世界的传感器噪声分布.
- PND可以灵活地处理不同噪声水平,减少建模的复杂性.
- PPM和DDL通过结合物理约束和明确的分配监督来提高合成噪声建模的准确性和精度.
- 在公共数据集上的实际低光原始图像消除任务中表现出卓越的性能.
结论:
- 拟议的基于暗的噪声建模策略显著打破了对训练无声化方法的数据依赖.
- PNNP提供了一种强大而高效的方法,用于近似真实世界的传感器噪声.
- 该方法显示了在推进实用的低光图像消光应用程序方面显著的潜力.
相关概念视频
Difference from Background: Limit of Detection
8.0K
The limit of detection (LOD) is the smallest amount of analyte that can be distinguished from the background noise. The LOD value corresponds to the concentration at which the analyte signal is three times larger than the standard deviation of the blank signal. Below this value, the analyte signal cannot be differentiated from the background noise. It is calculated by dividing the calibration slope by 3 times the standard deviation of the blank signals.
The LOD indicates the presence or absence...
The LOD indicates the presence or absence...
8.0K
Deconvolution
537
Deconvolution, also known as inverse filtering, is the process of extracting the impulse response from known input and output signals. This technique is vital in scenarios where the system's characteristics are unknown, and they must be inferred from the observable signals.
Deconvolution involves several mathematical techniques to derive the impulse response. One common approach is polynomial division. In this method, the input and output sequences are treated as coefficients of...
Deconvolution involves several mathematical techniques to derive the impulse response. One common approach is polynomial division. In this method, the input and output sequences are treated as coefficients of...
537
Light Acquisition
9.4K
In order to produce glucose, plants need to capture sufficient light energy. Many modern plants have evolved leaves specialized for light acquisition. Leaves can be only millimeters in width or tens of meters wide, depending on the environment. Due to competition for sunlight, evolution has driven the evolution of increasingly larger leaves and taller plants, to avoid shading by their neighbors with contaminant elaboration of root architecture and mechanisms to transport water and nutrients.
9.4K
