| 摘要 |
A gradient-based silhouette projection method for rendering the surface of a 3-D object is described herein. A set of imaging points is selected for imaging the object. The object is imaged from each of the selected imaging points to generate a set of images and a set of binary silhouettes are extracted from the images of the object, with at least one silhouette per image. The binary silhouettes are each filtered to generate an equal number of gradient silhouettes. Each of the gradient silhouettes is projected to generate a set of gradient volume projections that are combined to yield a single combined gradient volume. A 3-D isosurface is generated from the combined gradient volume, with the 3-D isosurface thereby representing the 3-D object. In one example described herein, the binary silhouettes are filtered to generate gradient silhouettes by determining tubular neighborhoods for the binary silhouettes. The gradient volume projections are represented by voxels. The 3-D isosurface is determined from the combined gradient volume by applying a marching cubes algorithm to the voxels representing the combined gradient volume. A set of polygons representing the 3-D isosurface is generated by approximately determining the set of polygons representing the locus of all lines that do not intersect the 3D isosurface. The set of polygons is subsequently textured based upon texture information provided in the set of input images or from other sources. The various steps may be repeated while the object moves or otherwise changes. A technique employing only a single initial silhouette is also described. Apparatus embodiments are also described.
|
