增强的CycleGAN网络与自适应暗通道,用于未配对的单图像脱
Yijun Xu1, Hanzhi Zhang2, Fuliang He2,3
1Westa College, Southwest University, Chongqing 400715, China.
Entropy (Basel, Switzerland)
|June 28, 2023
概括
这项研究引入了一种增强的CycleGAN模型,用于未配对的单图像除,通过调整暗通道前置 (DCP) 和优化修复过程以获得更清晰的图像来改进结果.
科学领域:
- 计算机视觉 计算机视觉
- 人工智能的人工智能
- 图像处理 图像处理
背景情况:
- 对于智能监控和遥感等应用来说,未配对的单图像脱雾至关重要.
- 现有的基于CycleGAN的方法存在人工痕迹和结果扭曲.
- 需要改进无监督的除烟技术.
研究的目的:
- 提出一种新的,增强的CycleGAN网络,用于未配对的单图像脱.
- 为了解决当前CycleGAN在图像消毒方面的局限性.
- 为了提高 dehazed 图像的准确性和视觉质量.
主要方法:
- 使用Wave-Vit语义细分模型进行自适应暗通道预先 (DCP) 估计.
- 从物理计算和随机抽样中优化了利用散射系数的重整过程.
- 通过大气散射模型,在一个增强的CycleGAN框架内集成排气和改造循环.
主要成果:
- 在基准数据集上实现了高性能:SSIM的94.9%和SOTS-outdoor上的PSNR的26.95.
- 在O-HAZE数据集上获得的SSIM为84.71%和PSNR为22.72.
- 在定量和视觉评估中,在现有算法上取得了显著的改进.
结论:
- 拟议的增强型CycleGAN与自适应的DCP有效地解决了未配对的单图像脱挑战.
- 该方法显著提高了图像质量,减少了工件和扭曲.
- 这种方法为现实世界的除尘应用提供了优越的解决方案.
相关概念视频
01:20Deconvolution
198
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...
198
02:16Light Acquisition
8.5K
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.
8.5K
01:19Masking and Demasking Agents
2.5K
EDTA titrations may necessitate masking and demasking agents to temporarily protect a particular metal ion in a mixture from the EDTA reaction. These agents facilitate the sequential analysis of the metal ions by forming stable complexes with some—but not all—metal ions during certain steps.
There are many masking agents, such as cyanide, fluoride, triethanolamine, thiourea, and 2,3-bis(sulfanyl)propan-1-ol (formerly 2,3-dimercapto-1-propanol), with the masking agent chosen based on...
There are many masking agents, such as cyanide, fluoride, triethanolamine, thiourea, and 2,3-bis(sulfanyl)propan-1-ol (formerly 2,3-dimercapto-1-propanol), with the masking agent chosen based on...
2.5K
01:26Phase Contrast and Differential Interference Contrast Microscopy
8.2K
Phase-Contrast Microscopes
In-phase-contrast microscopes, interference between light directly passing through a cell and light refracted by cellular components is used to create high-contrast, high-resolution images without staining. It is the oldest and simplest type of microscope that creates an image by altering the wavelengths of light rays passing through the specimen. Altered wavelength paths are created using an annular stop in the condenser. The annular stop produces a hollow cone of...
In-phase-contrast microscopes, interference between light directly passing through a cell and light refracted by cellular components is used to create high-contrast, high-resolution images without staining. It is the oldest and simplest type of microscope that creates an image by altering the wavelengths of light rays passing through the specimen. Altered wavelength paths are created using an annular stop in the condenser. The annular stop produces a hollow cone of...
8.2K