Blind Deconvolution for Poissonian Blurred Image With Total Variation and L₀-Norm Gradient Regularizations.

Area of Science:

• Image Processing
• Computer Vision
• Signal Processing

Background:

• Image deblurring is crucial for various applications.
• Poissonian noise and blur significantly degrade image quality.
• Existing blind deconvolution methods struggle with Poissonian noise.

Purpose of the Study:

• To develop a regularized blind deconvolution method for Poissonian blurred images.
• To improve image restoration quality and robustness against noise.
• To propose an efficient algorithm for solving the deconvolution problem.

Main Methods:

• Formulation using L0-norm for latent image and Total Variation (TV) for Point Spread Function (PSF) regularization.
• Incorporation of negative logarithmic Poisson log-likelihood.
• Variable splitting and Lagrange multiplier methods for solving the optimization problem.
• Alternating minimization algorithm for joint estimation of latent image and PSF.
• A non-blind deconvolution method with TV regularization for further enhancement.

Main Results:

• The proposed method successfully restores high-quality images from Poissonian blurred inputs.
• Experimental results demonstrate superior performance compared to state-of-the-art methods.
• The method is effective on both synthetic and real-world datasets.
• Achieved competitive or better results than existing techniques.

Conclusions:

• The proposed regularized blind deconvolution method is effective for Poissonian blurred images.
• The combination of L0-norm, TV regularization, and advanced optimization techniques yields high-quality restorations.
• This work offers a significant advancement in blind image deconvolution for noisy and blurred imagery.