Echo State Networks With Orthogonal Pigeon-Inspired Optimization for Image Restoration

Area of Science:

• Artificial Intelligence
• Computer Vision
• Signal Processing

Background:

• Image restoration is crucial for recovering original images from degraded versions (blurred or noisy).
• Neural networks offer a powerful framework for solving image restoration as a mapping problem.
• Echo State Networks (ESNs) are recurrent neural networks known for their simplified training.

Purpose of the Study:

• To propose a novel neurodynamic approach for image restoration.
• To enhance Echo State Network (ESN) performance through optimized parameter selection.
• To validate the effectiveness of the proposed method on various degraded images.

Main Methods:

• Utilizing an Echo State Network (ESN) for image restoration.
• Employing Pigeon-Inspired Optimization (PIO) to determine optimal ESN parameters.
• Incorporating an orthogonal design strategy in PIO for improved individual diversity.

Main Results:

• The proposed method demonstrates superior performance in restoring images with diverse blur and noise levels.
• Experimental results show improved image restoration quality compared to state-of-the-art techniques.
• The orthogonal PIO algorithm proves more effective than other bio-inspired optimization algorithms.

Conclusions:

• The neurodynamic approach with an optimized ESN provides effective image restoration.
• Orthogonal Pigeon-Inspired Optimization significantly enhances ESN training for image restoration tasks.
• This method offers a promising solution for recovering degraded images.