Camera Calibration for 3D Reconstruction and View Transformation - 3D Modeling and Animation: Synthesis and Analysis Techniques for the Human Body - page 123

calibration of other camera parameters. On the other hand, because in the
telepresence application the subsequent view reconstruction processing needs
the value of each individual camera parameter, all parameters have to be
calibrated explicitly.
Despite its wide range of applicability and extensive research, camera calibra-
tion is a subject for which there still remains some interesting and important
issues to be investigated, some of which are:
Optimal calibration pattern: For passive camera calibration, a calibration
target containing some control points with known geometry is needed. For
highly accurate calibration, the projections of these control points in the
image should be very efficiently and accurately located. Therefore, the
pattern of the control points should be carefully designed. In the past,
several kinds of patterns (e.g., circular dots and intersections of lines) have
been adopted. However, it is still not clear which existing pattern is the best.
Clear understanding of camera parameter mutual relation: Very little
attention has been paid to the analysis of the mutual dependencies among
all camera parameters. Due to the existing coupling, all camera parameters
should not be treated independently when one describes a 3-D scene. If
their mutual dependency and relative importance is clear, the more impor-
tant parameters can be treated “differently” during calibration. On the
other hand, the analysis of the uncertainty of each camera parameter could
also lead to better calibration algorithms.
Multiple camera configuration: The multiple camera configuration becomes
more and more popular in vision-based applications (Pedersini, Sarti, &
Tubaro, 1999), such as telepresence (Xu et al., 2002) and 3-D visual servo
systems (Stavnitzky & Capson, 2000). However, the passive calibration
problem of a multi-camera set-up has rarely been addressed. Self-calibra-
tion, on the other hand, always concentrates on multi-camera (or equiva-
lent) configurations. Therefore, passive calibration may borrow certain
techniques from self-calibration, especially for recovering the multi-cam-
era set-up. Similarly, self-calibration can benefit from passive calibration
on topics such as, the handling of nonlinear distortions.
References
Abdel-Aziz, Y. & Karara, H. (1971). Direct linear transformation into object
space coordinates in close-range photogrammetry. In Proc. symposium on
close-range photogrammetry , Urbana, IL, 1-18.