A stratified approach for camera calibration using spheres.

Area of Science:

• Computer Vision
• Geometric Calibration

Background:

• Traditional camera calibration relies on specific geometric features.
• Epipolar tangents on spheres are used for estimating epipolar geometry.
• Existing methods have limitations in feature extraction for precise calibration.

Purpose of the Study:

• To propose a stratified approach for camera calibration using spheres.
• To introduce novel point features derived from bitangent envelopes of spheres.
• To develop an algorithm for recovering the fundamental matrix and absolute dual quadric for full camera calibration.

Main Methods:

• Exploiting bitangent envelopes of sphere pairs to identify two additional point features.
• Developing a method for locating these point features and sphere centers in images.
• Formulating the absolute dual quadric as a cone tangent to a dual sphere.
• Utilizing recovered image points and epipolar tangents from three spheres to recover the fundamental matrix.
• Upgrading weak calibration to full calibration using the recovered absolute dual quadric.

Main Results:

• Successfully identified and utilized new point features from spheres for camera calibration.
• Developed a robust algorithm for fundamental matrix recovery in a plane plus parallax representation.
• Derived a novel formulation for the absolute dual quadric.
• Demonstrated high precision and feasibility through experiments on synthetic and real data.

Conclusions:

• The proposed stratified approach using spheres offers a precise and feasible method for camera calibration.
• The introduction of new point features and the absolute dual quadric formulation significantly advance calibration techniques.
• The algorithm effectively upgrades weak calibration to full calibration, validated by experimental results.