MSF-ACA: Low-Light Image Enhancement Network Based on Multi-Scale Feature Fusion and Adaptive Contrast Adjustment

Area of Science:

• Computer Vision
• Image Processing
• Deep Learning

Background:

• Existing low-light image enhancement methods struggle with detail loss, poor contrast, and high computational demands.
• These limitations hinder the practical application of image enhancement technologies in various fields.

Purpose of the Study:

• To develop an efficient and robust low-light image enhancement network (MSF-ACA).
• To address the challenges of detail preservation, contrast enhancement, and computational complexity in low-light imaging.

Main Methods:

• The proposed MSF-ACA network utilizes multi-scale feature fusion and adaptive contrast adjustment.
• Key components include the local-global image feature fusion module (LG-IFFB) and the adaptive image contrast enhancement module (AICEB).
• LG-IFFB employs a dual-branching structure for multi-scale feature extraction and fuses local details with global illumination. AICEB dynamically adjusts computational depth based on feature map confidence.

Main Results:

• The MSF-ACA network has a low parameter count (0.02 M).
• Achieved 21.53 dB PSNR on the LOL-v2-real dataset and a BRI of 16.04 on the DICM dataset.
• Demonstrated superior detail clarity and color fidelity compared to mainstream algorithms.

Conclusions:

• The MSF-ACA network offers a highly efficient and robust solution for low-light image enhancement.
• It effectively balances contrast enhancement and computational efficiency while preserving crucial image details.
• The proposed method significantly improves visual quality in challenging low-light conditions.