Camera Calibration for 3D Reconstruction and View Transformation - 3D Modeling and Animation: Synthesis and Analysis Techniques for the Human Body - page 71

Camera calibration is the process of determining the internal physical character-
istics of a camera and its 3-D position and orientation with respect to a world
coordinate system using some predefined objects or automatically detected
features. The result of camera calibration is the establishment of a mathematical
relationship between the 3-D coordinates of a point in the 3-D scene and the 2-D
coordinates of its projection onto the image recorded by the camera.
Camera calibration is an important preliminary step towards many vision-related
applications. Passive calibration, active calibration, and self-calibration are the
most frequently referred to camera calibration algorithm categories. Active
calibration methods were developed mainly for robotic systems. Recently,
algorithms for active calibration purposes have been investigated that fall in the
more general self-calibration category (Lamiroy, Puget & Ho-raud, 2000).
While detailed discussions about self-calibration are given in Faugeras & Luong
(2001), Hartley & Zisserman (2000) and Fusiello (2000), this paper intends to
give an overview of passive calibration. However, for completeness, a brief
counting about the self-calibration will also be presented.
Passive calibration has been used extensively in the synthesis and analysis of the
human body for telepresence (Xu, Lei, & Hendriks, 2002) and in 3-D face
modeling (Liu, Zhang, Jacobs, & Cohen, 2001). However, despite its wide range
of applications and extensive investigations, no comprehensive overview of this
topic exists. This chapter attempts to fill this gap by providing such an overview
in a systematic and unified manner and by comparing and evaluating existing
approaches. In addition, two typical applications are given of passive camera
calibration methods for specific problems of face model reconstruction and
telepresence and then experimentally evaluated. It is expected that this chapter
can serve as a standard reference. Researchers in various fields in which
passive camera calibration is actively or potentially of interest can use this
chapter to identify the appropriate techniques suitable for their applications.
The chapter is organized as follows. In the next section, a detailed introduction
and mathematical description is provided of the imaging process of an off-the-
shelf camera. In the next section, all existing camera calibration techniques are
classified based on several different points of view. The nonlinear component
of the camera, responsible for distortion, is then modeled using two alternative
methods and discussed in a following section. Key passive camera calibration
algorithms are reviewed in detail, followed by a brief overview of self-calibration
algorithms. Finally, two applications for which calibrated cameras are required
are analyzed, and a summary and conclusions are presented.