Automatic Tuning of Gaussian Filter for Image Vignetting Correction.

Area of Science:

• Optics and Image Processing
• Computational Imaging

Background:

• Image vignetting is a common optical artifact reducing image quality and radiometric accuracy.
• Existing vignetting correction methods often use Gaussian filtering but struggle with subjective parameter selection.

Purpose of the Study:

• To develop an automated method for vignetting correction.
• To improve the accuracy and reliability of vignetting estimation and image brightness uniformity.

Main Methods:

• Proposed the Gaussian filter with auto-tuned sigma (GFATS) method.
• Utilized an optimization framework to automatically tune Gaussian filter parameters by aligning filtered output with a polynomial model of vignetting.
• Derived vignetting correction matrix using the tuned filter.

Main Results:

• GFATS provides more accurate vignetting estimation and improved brightness uniformity compared to existing methods.
• The method effectively mitigates overfitting issues common in standard smoothing techniques.
• A single, consistent parameter set demonstrated reliable performance across diverse lens-camera systems.

Conclusions:

• GFATS offers a versatile and practical solution for computational image vignetting correction.
• The automated parameter tuning enhances accuracy and robustness in various imaging applications.
• The proposed method represents a significant advancement over traditional vignetting correction techniques.