The CCD (Charge Coupled Device) image formation theory is at the foundation of 3D vision systems. Ideally, a five-element block diagram can model this process. The first block represents the nonlinear distortions caused by camera lenses. The second and third elements gather the low-pass filter effects due to lens aberrations and CCD phenomenon. A fourth block illustrates the quantization effects induced by a series of discrete photosensitive elements on the CCD and by the A/D conversion for analog cameras. The last block represents the addition of random noise on the discrete signal. Because step-like luminance transitions undergoing lens distortion remain step-like, it is possible to precisely correct for the distortions after the edge localization. The efficient distortion correction process is exactly where the camera lens calibration challenge resides. The calibration exercise also seeks the intrinsic and extrinsic camera parameters, i.e. the information that relates to the camera optics and the information that describes the location and orientation of the camera in 3D space. This thesis presents a review and evaluation of several methods designed for optimal accuracy on the parameters evaluation. (Abstract shortened by UMI.)
展开▼
