COMMENT ⊗   VALID 00002 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	THE STATE OF RESEARCH IN COMPUTER VISION
C00006 ENDMK
C⊗;
THE STATE OF RESEARCH IN COMPUTER VISION


	The  use of  computers  to process  pictures  dates from  the
1950's and has  many aspects ranging from simple procedures to reduce
noise and increase contrast to work in artificial  intelligence aimed
at  finding out  what is  going  on in  a  complex scene.   We  shall
discuss  the  state  of  this research  from  the  point  of  view of
identifying aspects  of it  that might have  defense applications  in
the next two to ten years.

HISTORICAL

1. local operators

2. perceptrons and other discrimination schemes

3. About  1960 or 1961 McCarthy  and Minsky proposed  work on systems
that describe scenes rather than classify them.

4. Roberts

5. Systematic work on scene description and its use for  manipulation
started in the M.I.T. Artificial Intelligence  Laboratory in 1965 and
in the Stanford  Artificial Intelligence Laboratory in 1966.  In that
time M.I.T.  has  produced about  xx  reports and  PhD  dissertations
dealing with computer vision and Stanford has produced about yy.

6.  More recently  many  other laboratories  have  entered the  scene
description field  some of them starting from the pattern recognition
side and working towards scene description.

7. Stanford University has tended to emphasize the  low level aspects
of  vision  including   edge  finders,  region  growers,  correlation
techniques.   Recently, much  of this  work has  concentrated on  the
three   dimensional   aspects   of    outdoor   scenes.   Forthcoming
dissertations  by Yakimovsky  (adviser =  Jerome Feldman)  and Hannah
(adviser = John McCarthy) emphasize different aspects of this.

8. M.I.T.  has emphasized higher  level aspects  of vision  including
planning,

9. Stanford Research Institute (not  the same as Stanford University)
has emphasized the use of relatively simple techniques to acquire
economically and reliably certain specific types of information
from visual data.