Nonlinear Deblurring for Low-Light Saturated Image.

Area of Science:

• Computer Vision
• Image Processing

Background:

• Single image deblurring excels for natural images.
• Saturation in blurry images arises from low light and long exposures.
• Linear deblurring methods cause ringing artifacts on saturated, low-light images.

Purpose of the Study:

• To address limitations of linear deblurring for saturated, low-light images.
• To develop a deblurring method that reduces artifacts and estimation errors in saturated regions.
• To propose a nonlinear model for adaptive pixel saturation deblurring.

Main Methods:

• Formulated saturation deblurring as a nonlinear model with adaptive pixel handling.
• Introduced a nonlinear function to the convolution operator for saturation.
• Utilized a maximum-a posteriori framework and the alternating direction method of multipliers (ADMM) for efficient computation.

Main Results:

• The proposed method reduces estimation errors in saturated areas and suppresses ringing artifacts.
• It achieves high-quality restoration comparable to conventional deblurring methods.
• The nonlinear degradation model simplifies saturation and unsaturated degradation formation without complex detection.

Conclusions:

• The nonlinear deblurring approach effectively handles low-light saturated images.
• It outperforms state-of-the-art methods in both synthetic and real-world scenarios.
• The method offers a straightforward and robust solution for saturation deblurring.