An algorithm for constructing the octree representation of a 3-D object from silhouette images of the object is described. The images must be obtained from 13 viewing directions corresponding to the three 'face' views, six 'edge' views, and four 'corner' views of an upright cube. These views were chosen because they provide a simple relationship between the pixels in the image and the octant labels in the octree. The accuracy of the constructed octree is defined to be the ratio of the object volume to the octree volume. Several geometric objects placed in random orientations were used to test the accuracy of the octree generation algorithm. The average accuracy was found to be greater than 90%. The variation in accuracy with changes in object orientation was found to be fairly low. The accuracy and its dependence on viewpoint were also tested for a more complex object composed of simple primitives. Both accuracy and its dependence on object orientation were found to be only marginally lower for the complex object.
|Original language||English (US)|
|Title of host publication||Unknown Host Publication Title|
|Number of pages||6|
|State||Published - Jan 1 1986|
ASJC Scopus subject areas