Multiresolution reconstruction of pipe-shaped objects from contours

We reconstruct pipe-shaped objects from a set of contours, each of which is extracted from an image representing a slice sampled from 3D volume data. The contours are formed by connecting the intersection points between rays cast from a central pixel of an image slice and the boundary of the shape. The edges on the contours are classified into several types, which are exploited in triangulating the contours, thus eliminating most of the floating-point computation from the tiling. Initially, contours of lowest resolution are extracted to reconstruct a lowest-resolution object, which is refined by adding points to the contours.