Turbulent-PSO-Based Fuzzy Image Filter With No-Reference Measures for High-Density Impulse Noise

Area of Science:

• Digital Image Processing
• Computational Intelligence
• Signal Processing

Background:

• Digital images are susceptible to impulsive noise during acquisition, transmission, and processing.
• Existing noise reduction methods often require noise-free reference images or struggle with high noise rates.
• Genetic-based algorithms have been explored for image noise removal but have limitations.

Purpose of the Study:

• To propose a novel turbulent particle swarm optimization (TPSO)-based fuzzy filtering (TPFF) approach.
• To effectively remove impulse noise from highly corrupted digital images.
• To develop a filter that does not require noise-free reference images for parameter optimization.

Main Methods:

• Implementation of a fuzzy filter comprising parallel fuzzy inference, fuzzy mean, and fuzzy composition processes.
• Optimization of filter parameters using TPSO, leveraging its success rate and solution quality.
• Utilization of a no-reference Q metric for TPSO learning, enabling online adaptation and parameter tuning.

Main Results:

• The TPFF demonstrates superior performance in impulse noise removal compared to existing methods.
• Restored images achieve excellent quality, validated by peak signal-to-noise ratio (PSNR), mean square error (MSE), and mean absolute error (MAE).
• Effective noise reduction is achieved even at noise rates exceeding 0.5, without requiring ground-truth images.

Conclusions:

• The proposed TPFF is an effective and improved online version of noise filtering algorithms.
• TPSO-based optimization allows the filter to adapt and perform well in practical scenarios without reference images.
• The TPFF offers a robust solution for restoring highly corrupted digital images, significantly outperforming previous techniques.