COMMENT ⊗   VALID 00005 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	5.0	HIDDEN LINE ELIMINATION FOR COMPUTER VISION.
C00003 00003	5.0	Introduction.
C00004 00004	5.1	Hiding a Topologically Coherent Object.
C00006 00005	
C00009 ENDMK
C⊗;
5.0	HIDDEN LINE ELIMINATION FOR COMPUTER VISION.

5.0	Introduction.
5.1	Hiding a Topologically Coherent Object.
5.2	Two simple hidden line eliminators that almost work:
		Edge-Edge and Face-Vertex.
5.3	2-D Face, Edge, Vertex Sorting.
5.X	Propagating Underfaces.
5.X	Shadows.
5.X	Photometric Modeling and Video Generation.
5.X	Performance
5.0	Introduction.

	Hidden line elimination  refers to the process  of simulating
the  appearance  of  a  collection  of  opaque objects.  In  computer
graphics the results  of a hidden line  elimination are intended  for
humans, while in  computer vision the output is  intended for further
machine   processing,  namely  for  the  image  comparison  which  is
verification vision.

Three reasons why
hidden line elimination is a perenial research problem
that has yet been adequately solved

solutions are quite lacking in both realism, economy and flexibility.
5.1	Hiding a Topologically Coherent Object.

	OCCULT marks  the faces, edges  and vertices of  a polyhedral
scene  as either visible  or hidden with  respect to  a camera model.
Edges that were at first  partially visible are split into pieces  so
that each piece is either fully visible or fully hidden.

Like Warnock's  algorithm, OCCULT  has two  parts: a  recursive outer
part that splits an image until there are only afew edges in it or it
has  become too  small;  the  inner  part is  the  real  hidden  line
eliminator.  The  inner  part of  OCCULT  is  based  on the  idea  of
comparing each edge with  all the other edges  and hiding as much  as
possible when a pair of intersecting edges is discovered.

	OCCULT is a hidden line eliminator; it is neither  a  Watkins
nor  a Warnock algorithm but is rather a throw-back to the naive idea
of comparing each edge with all the other edges and  having  ways  to
dampen the potentially large number of comparisons that might occur.

	There are three kinds of dampening in OCCULT. The first (used
in other hidden eliminators) is to get rid of  the  faces  that  have
their  backs  to  the camera and to consider for comparision only the
edges with one potentially visible face.    These  edges  are  called
"folds".  The  second  kind  of  dampening,  is  to  hide  everything
connected to the hidden portion of an edge when a  fold  crossing  is
discovered, this is made possible by the winged edge primitives which
allow polyhedron surfaces to be easily traversed  topologically;  and
by  the  Euler primitives which allows the edges to be quickly broken
into visible and hidden portions without losing their  topology.  The
third  kind  of dampening involves having a raster of edge buckets to
localize the comparisons.

	The reason for doing hidden line elimination in this  fashion
is  to  get  the topology of the image regions and edges in a modeled
scene including the shadows. OCCULT was used  to  make  some  of  the
figures  that  appeared  earlier  in  this paper; for example the arm
model in figure 1.2, (which required twelve seconds of PDP-10 compute
time).