Geometric calibration of central omnidirectional cameras under the unit-viewing sphere model
Year: 2026
Type: article
Abstract: This paper presents a full-parameter calibration algorithm for central omnidirectional cameras, including central catadioptric and central fisheye cameras, unified under the unit-viewing sphere model. Existing methods typically require multiple line or chessboard patterns and are sensitive to radial distortion and image noise. In contrast, the proposed algorithm selects line images and views to identify those that provide valid calibration information. Building on this analysis, it requires only a mirror contour and one valid line image from a single view to linearly recover all intrinsic and spherical projection parameters. When at least four planar-point correspondences (or six world-point correspondences) are available, the extrinsic parameters and distortion coefficients can also be estimated. Notably, the effect of radial distortion can be approximately compensated by estimation biases in the effective focal length and the spherical projection parameter. Experimental results demonstrate that the proposed algorithm achieves robustness and accuracy comparable to classical methods, while enabling the selection of geometrically valid constraints.
Source: Optics Express
DOI: 10.1364/oe.613936
Authors: Zhuo Qing-Feng, Zhao Yue, Li Chao-Qian, Jiang Nan
Language: English
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: gold
Estimated APC paid: $2,270
OpenAlex ID: W7223684191