G. Barequet and A. Zamansky (Israel)
Surface interpolation, contour correspondence.
This paper presents a practical solution for three dimensional shape reconstruction from sparsely sampled two-dimensional data, represented as contours on a set of parallel slices. Our method separates the solution for con tour correspondence from the surface tiling, and is guar anteed to produce non-self-intersecting surfaces regardless of the contours’ geometry, complexity, and the computed correspondence. Though our algorithm is fully automatic, human intervention is also possible for refinement. First, we apply a method based on object recogni tion techniques to find contour correspondence. Next, we use Voronoi diagrams and operations on a planar map to create a partition, which guarantees a correct reconstruc tion. Finally, a surface-interpolation algorithm is applied to the contours translated according to both their corre spondence and the created partition to produce the actual surface. Experimental results demonstrate the good perfor mance of our algorithm and improvement in the quality of reconstruction for sparsely-sampled inputs.
Important Links:
Go Back
