FSD-BRIEF: A Distorted BRIEF Descriptor for Fisheye Image Based on Spherical Perspective Model

Area of Science:

• Computer Vision
• Image Processing
• Robotics

Background:

• Fisheye images possess a wide Field of View (FOV) but suffer from severe radial distortion.
• Traditional feature descriptors perform poorly with distorted fisheye images, hindering accurate feature matching.

Purpose of the Study:

• To develop a novel descriptor for robust feature matching in fisheye images.
• To address the limitations of existing methods in handling significant radial distortion.

Main Methods:

• A distorted Binary Robust Independent Elementary Feature (BRIEF) descriptor was designed based on a spherical perspective model.
• A 3D gray centroid of feature points was utilized, and a feature point attitude matrix described their position and orientation.
• A coordinate mapping relationship was established between the BRIEF descriptor template and the fisheye image for distorted descriptor computation.

Main Results:

• The proposed descriptor demonstrated effective distortion invariance.
• Experimental results showed a significant improvement in feature matching performance for fisheye images.
• The descriptor accurately handles the radial distortion inherent in wide FOV imagery.

Conclusions:

• The novel distorted BRIEF descriptor offers a robust solution for feature matching in fisheye images.
• This method enhances the reliability of computer vision systems utilizing wide-angle lenses.
• The approach significantly improves upon traditional feature descriptors in distorted image scenarios.