Improved vector-extrapolation-based Richardson-Lucy algorithm used for wavefront coded imaging.

Area of Science:

• Image Restoration
• Computational Imaging
• Signal Processing

Background:

• The Richardson-Lucy (RL) algorithm is a standard nonlinear method for image restoration, particularly in astronomy and microscopy.
• Classical RL algorithms exhibit slow convergence and limitations with wavefront coded images, hindering real-time applications.
• Existing vector-extrapolation RL methods face challenges with noise amplification and require post-processing.

Purpose of the Study:

• To develop an improved Richardson-Lucy algorithm for efficient and high-quality restoration of wavefront coded blurry images.
• To accelerate convergence and mitigate noise amplification inherent in existing RL-based methods.
• To introduce an adaptive stopping criterion for optimal iteration count per color channel.

Main Methods:

• Introduction of an exponential modified correction term into the vector-extrapolation-based RL algorithm.
• Development of a structure-similarity-index-metric (SSIM)-based stopping criterion for adaptive iteration control.
• Application of the improved algorithm to wavefront coded blurry images.

Main Results:

• Achieved a significant reduction in total iterations by approximately 78.9%.
• Effectively prohibited noise amplification, leading to improved signal-to-noise ratio.
• Restored images exhibited superior visual quality without the need for additional denoising.

Conclusions:

• The proposed improved RL algorithm offers a faster and more effective solution for restoring wavefront coded images.
• The integration of the correction term and SSIM-based stopping criterion enhances both speed and quality.
• This advancement holds promise for real-time applications in image restoration fields.