Non-dyadic fisheye lens correction model for image enhancement

Area of Science:

• Computer Vision
• Image Processing
• Optics

Background:

• Radially distorted images from wide-angle lenses often suffer from jagging and blurring artifacts.
• Existing methods for fisheye lens distortion correction may not adequately address these artifacts.

Purpose of the Study:

• To present a novel non-dyadic framework for enhancing example-based correction of radially distorted images.
• To improve the correction of barrel distortion from fisheye lenses, minimizing artifacts.

Main Methods:

• Employs a non-dyadic (multiple-step) geometric correction based on a parabolic equation lens distortion model.
• Integrates an example-based image enhancement technique at each correction step to mitigate artifacts like jagging and blurring.

Main Results:

• The proposed method effectively removes jagging and blurring artifacts during fisheye lens distortion correction.
• Experimental results show superior performance compared to existing fisheye lens image enhancement methods based on subjective and objective evaluations.

Conclusions:

• The developed non-dyadic framework significantly enhances the quality of corrected wide-angle images.
• This method is suitable for diverse wide-view imaging applications, including automotive cameras and surveillance systems.