EPIPOLAR PARAMETERIZATION FOR RECONSTRUCTING 3D RIGID CURVE

This paper describes a new method for reconstructing a 3D rigid curve from a sequence of uncalibrated images using 3D epipolar parameterizat ion. The approach can be divided into the following two steps: First, a nonlinear discrete method is presented for point by point reconstruc tion of the curve instead of whole curve reconstruction. Projective an d Euclidean geometric tools are used. Second, the parametric represent ation of the curve is defined by 3D B-spline curves. A linear method i s proposed to interpolate the reconstructed points to obtain a complet e curve. Thus it is proved that the 3D curve interpolation is equivale nt to determining a set of control points of 3D regularized B-spline c urves. It is shown that the 3D epipolar parameterization is an efficie nt method for reconstructing a 3D curve from image sequences. Experime ntal results are presented for real data. (C) 1997 Pattern Recognition Society. Published by Elsevier Science Ltd.