الفهرس 
AbstractOnly 14 pages are availabe for public view
 |  | from | 229 |  |  |
| | | from | 229 | | |
Abstract
The need for efficient 3D object representations is crucial in Computer Vision , Computer Graphics , Computer Aided Design and other related areas. Octrees are used in many 3D representation problems because they provide compact data structures , allow rapid access to occupancy information • and implement many geometric manipulation algorithms efficiently.
In this thesis , the following problem is addressed : Given a set of silhouette views of an object, construct the Octree representirig the object which gave rise to those views . A new algorithm to construct the octree representation of an object from silhouette images taken from any viewpoint is described. We then compare its performance with the most efficient algorithm available in the literature and prove that it has a better time and space complexity and a slightly better accuracy .
The issue of Octree representation and processing is addressed next. The existing representation techniques are reviewed, then three new representation approaches are introduced. Different representations are then compared with respect to ability to store large amounts of data • ability to skip details , compactness , and ease of processing . It is concluded from the analysis and experimental results that one of the new representations has the best performance on all accounts .