These types of images do not meet the requirements of advanced visual tasks, so low-light image enhancement is currently a flourishing and challenging research topic. To alleviate the problem of low brightness and low contrast, this paper proposes an improved zero-shot Retinex network, named IRNet...
在本文中,我们提出了一种新的零参考低照度增强框架,该框架可仅使用正常光照下的图像进行训练。为此,我们从物理光传递理论中汲取灵感,设计了一种光照不变先验。这个先验值是连接正常光线图像和弱光图像的桥梁。然后,我们开发了一个先验图像框架,在没有弱光数据的情况下进行训练。在测试过程中,该框架能够将我们的光照...
首先使用CLIP图像编码器Φimg和使用CLIP文本编码器Φtxt学习的提示对增强图像ˆI进行编码。接下来,计算嵌入增强图像 Φimg ( ^I) 和学习提示对 Φtxt(P ) 之间的余弦相似度,其中 P = Pp, Pn 是一对正负学习提示。 (2)语义指导 利用CLIP 模型的 zeroshot 能力,以直接的方式在训练期间引入语义指导。为了简...
In [11], the light enhancement is creatively formulated as a task of image-specific curve estimation us- ing zero-shot learning. In [20, 47, 55], 3D lookup table and color histogram are utilized to preserve the color consis- tency. However, existing designs...
For low-light enhancement tasks, network architecture design needs to be adapted to the characteristic of low-light images having more low-light features than high-light features. At the same time, for low-light enhancement tasks in real scenes, zero-shot learning32 methods are needed to ...
computer-visiondetectiontransformerimage-restorationlow-light-enhancelow-light-image-enhancementlow-light-vision UpdatedApr 28, 2025 Python Pytorch implementation of Zero-Reference Deep Curve Estimation for Low-Light Image Enhancement deep-learningpytorchhdrzero-shot-learninglow-light-enhancelow-light-image-enh...
Existing low-light image enhancement techniques face challenges in achieving high visual quality and computational efficiency, as well as in effectively removing noise and adjusting illumination in extremely dark scenes. To address these problems, in thi
Zhang Y, Guo X, Ma J, Liu W, Zhang J (2021) Beyond brightening low-light images. Int J Comput Vision 129:1013–1037 Article Google Scholar Zheng S, Gupta G (2022) Semantic-guided zero-shot learning for low-light image/video enhancement. In: Proceedings of the IEEE/CVF Winter Confere...
To drive the zero-shot learning, a combination of Retinex reconstruction loss, texture enhancement loss, and illumination-guided noise estimation loss is proposed. Zhao et al. [31] perform Retinex decomposition via neural networks and then enhance the low-light image based on the Retinex model, ...
Zero-DCE方法:定义无参考损失函数Non-Reference Loss Functions,不需要任何paired或unpaired的数据 二、算法实现 Zero−DCE包括如下三个部分: Light-Enhancement curves (LE-curves) 光增强曲线 Deep Curve Estimation Network(DCE-Net) 深度曲线估计网络 Non-Reference Loss 无参考损失函数 ...