PSO-based parameter optimization of intuitionistic fuzzy generator for low-light image enhancement

Area of Science:

• Computer Vision
• Image Processing
• Artificial Intelligence

Background:

• Low-light conditions degrade image quality, causing reduced visibility, noise, and loss of structural details.
• These degradations negatively impact visual perception and the accuracy of subsequent image analysis tasks.
• Existing methods may require extensive training data or fail to adapt to diverse low-light scenarios.

Purpose of the Study:

• To develop an adaptive low-light image enhancement framework.
• To improve contrast and preserve structural details in images captured under insufficient illumination.
• To provide an interpretable enhancement solution suitable for scenarios lacking ground-truth data.

Main Methods:

• A preprocessing step using Block-Matching and 3D filtering (BM3D) for noise reduction while preserving structure.
• An Intuitionistic Fuzzy Generator (IFG) to model pixel intensity uncertainty for adaptive contrast enhancement.
• Gamma correction for brightness adjustment, with parameters optimized via Particle Swarm Optimization (PSO) using SSIM or entropy-based metrics.

Main Results:

• The proposed framework demonstrated competitive enhancement performance on benchmark datasets.
• Significant improvements in image contrast and preservation of visually relevant details were observed.
• Effective performance was achieved in both reference (ground-truth available) and no-reference scenarios.

Conclusions:

• The IFG-based framework offers an effective approach to low-light image enhancement, particularly when training data is unavailable.
• The parameter-adaptive nature, guided by PSO, allows for tailored enhancement based on image content and quality metrics.
• While computationally more intensive than deep learning models, it provides interpretable and adaptable solutions for specific applications.