Related Experiment Video
Updated: Aug 2, 2026
08:32
Optimized Protocol for Retinal Wholemount Preparation for Imaging and Immunohistochemistry
Published on: December 13, 2013
23.3K
A depth iterative illumination estimation network for low-light image enhancement based on retinex theory
Yongqiang Chen1, Chenglin Wen2,3, Weifeng Liu4
1School of Automation, Guangdong University of Petrochemical Technology, Maoming, 525000, China.
Scientific Reports
|November 12, 2023
Summary
This study introduces a novel illumination enhancement network using Retinex theory to improve low-light images. The method effectively enhances brightness, suppresses noise, and preserves details, outperforming existing techniques.
Area of Science:
- Computer Vision
- Image Processing
- Artificial Intelligence
Background:
- Low-light image enhancement faces challenges in visual quality, computational efficiency, noise removal, and illumination adjustment.
- Existing methods struggle with extremely dark scenes, necessitating advanced solutions.
Purpose of the Study:
- To propose a fast and accurate illumination enhancement network based on Retinex theory for low-light images.
- To address limitations in visual quality, noise, and illumination adjustment in existing techniques.
Main Methods:
- Developed a two-stage learning-based network: a decomposition network and an enhancement network.
- The decomposition network separates images into reflectance and illumination maps.
- The enhancement network uses illumination enhancement and reflection denoising modules, employing cascaded iterative learning and weight sharing for accurate estimation.
Main Results:
- The proposed framework effectively suppresses noise and preserves details in low-light images.
- Achieved a 9.16% increase in Peak Signal-to-Noise Ratio (PSNR) over Retinex-Net on the LOL dataset.
- Demonstrated a 19.26% improvement compared to the state-of-the-art SCI method.
Conclusions:
- The novel illumination enhancement network offers superior performance in low-light image processing.
- The method provides significant improvements in brightness, noise reduction, and detail preservation.
- Unsupervised training losses enhance the model's generalization capabilities for diverse low-light conditions.
Related Concept Videos
01:24Vision
Vision is the result of light being detected and transduced into neural signals by the retina of the eye. This information is then further analyzed and interpreted by the brain. First, light enters the front of the eye and is focused by the cornea and lens onto the retina—a thin sheet of neural tissue lining the back of the eye. Because of refraction through the convex lens of the eye, images are projected onto the retina upside-down and reversed.
02:16Light Acquisition
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.