DLFE-Net: Preserving Details and Removing Noise Using HVI Color Space for Low-Light Image Enhancement

Area of Science:

• Computer Vision
• Image Processing
• Artificial Intelligence

Background:

• Low-light image enhancement (LLIE) is crucial for various applications.
• Existing methods struggle with overexposure, detail loss, and noise in low-light conditions.

Purpose of the Study:

• To propose a novel network, DLFE-Net, for effective low-light image enhancement.
• To address challenges of local overexposure, detail preservation, and noise removal in LLIE.

Main Methods:

• The proposed Denoiser and Low-Frequency Enhancer Network (DLFE-Net) converts RGB to HVI color space.
• It utilizes a Low-Frequency Illumination Enhancer (LFIE) module for detail preservation and overexposure mitigation.
• A Multi-Scale Gated Denoiser (MSGD) module is employed for effective noise removal.

Main Results:

• DLFE-Net demonstrated superior performance over state-of-the-art methods on benchmark and unpaired datasets.
• Qualitative and quantitative analyses confirmed the effectiveness of the proposed approach.
• Ablation studies validated the contribution of individual modules within DLFE-Net.

Conclusions:

• DLFE-Net offers a robust solution for low-light image enhancement.
• The network successfully mitigates overexposure, preserves details, and removes noise.
• The proposed LFIE and MSGD modules are effective in addressing key LLIE challenges.