Efficient generalized cross-validation with applications to parametric image restoration and resolution enhancement

Area of Science:

• Image processing
• Computational imaging
• Inverse problems

Background:

• Image restoration and resolution enhancement often face challenges with unknown or partially known blurring processes (point spread functions, PSF).
• Accurately estimating the PSF and regularization parameters is crucial for solving ill-posed inverse problems in imaging.

Purpose of the Study:

• To develop a robust method for estimating unknown point spread function (PSF) parameters and regularization parameters in image restoration.
• To improve the efficiency and computational performance of parameter estimation for ill-posed inverse problems.

Main Methods:

• Utilized generalized cross-validation (GCV) for simultaneous estimation of PSF and regularization parameters.
• Implemented efficient approximation techniques based on the Lanczos algorithm and Gauss quadrature theory.
• Validated the data-driven approach using synthetic and real image sequences.

Main Results:

• Demonstrated effective and robust estimation of PSF and regularization parameters from raw image data.
• Significantly reduced the computational complexity of the GCV method through proposed approximations.
• Showcased the practical applicability of the method in image restoration tasks.

Conclusions:

• The proposed method provides an effective solution for estimating unknown PSF and regularization parameters in image restoration.
• Efficient approximation techniques enhance the feasibility of GCV for complex imaging problems.
• The data-driven approach offers a robust tool for improving image quality in various applications.