| Summary: | 碩士 === 淡江大學 === 資訊工程研究所 === 81 === A fundamental problem in computer vision is how to re-
coverhe three - dimensional configurations of a scene from its
image.he problem is generally under constrained and
complicated. Thisaper investigates computational mechan- ism
for interpretation ofine drawings by means of which a
machine can extract three -imensional object structures from
their pictures draws on a two -imensional plane. Given a
line - drawinng of a three -imensional scene, a human
being usually has definite ideas abouthe three - dimensional
shape of objects. For computers, linerawings are simple
collections of line segments on awo-dimensional
plane. In order to extract three - dimensionalnformation
from them, some intelligent mechanism is necessary.he
objects considered here are polyhedrons, that is,
solidbjects bounded by planar faces, and the line drawings are
singleview pictures of these objects. Probably this mechanism
is quiteifferent from what is employed in human
perception, given aingle two-dimensional picture of an object,
a human being usuallyas definite ideas about its three -
dimensional shape, in spitef the fact that there exist a
large number of shapes which canroduce the same picture.
This fact indicate that we use somessumptions or
knowledge about the objects and about imageormation. We
shall also investigate some of these assumptions andhen study
some emulation formulation. Under this formulation, theroblem
is to emulate the ability of the human vision system
tonterpret images as three- dimen- sional objects.
|
|---|
