From these result, we can find high definition images more 
than Figure 5 
Figure 1 was taken by a CCD camera which has 400,000 
pixels. At this investigation, the authors used new method 
to use a image which taken by mega-pixel CCD camera 
(using 1,000,000 pixel CCD, mega-pixel image). 
Figure 7 shows edged result with mega-pixel image. 
  
  
Figure 7(a) Edged image 
(Target 1, mega-pixel image) 
e . 
~ 
SS 
4 
^8 
ETE 
== 
NSS 
c 
2» 
2 
25. 
iie 
S 4 
NY 
2 
2. 
  
Figure 7(b) Edged image 
(Target 2, mega-pixel image) 
From these results, it is to be understood that though such 
an edge as to have been extracted even by the filtering 
technique is seen only in the height image, a noise part 
has been reduced greatly by taking account of a relation 
with the reliability and a better edged result than the 
conventional one has been secured. 
Especially, if we use high resolution image, we can get 
more minuteness edged image. 
5. CONCLUSION 
The employment of an edge technique (reliability method) 
using an edge height and reliability has made it possible to 
acquire an edged image which is superior to that obtained 
by the conventional LG filter in the respect of the detection 
performance of edge by enhancing mainly the reduction of 
noise in the image. 
The questions to be worked out in future are thought to be 
as follows: 
a. Automation of setting thresholds 
(At present, they are set manually.) 
b. Improvement in edge performance 
ACKNOWLEDGEMENTS 
The authors were got much advice from Mr. SUGIYAMA. 
The authors express graduate to Mr. SUGIYAMA for his 
help at this place. 
REFERENCE 
1.Hiroshi YOKOYAMA, Katsuhiro Hatano, Hirofumi 
CHIKATSU, 1996, Ortho Projection Drawing for 
Archeological Artifacts of complicated form, International 
Archives of Photogrammetry and Remote Sensing, Vienna, 
Vol. XXXI, PartB5,pp.95-100 
2.Devid MARR, 1982.VISION, W. H. Freeman and 
Company 
3.Takahiro SUGIYAMA, Keiichi ABE, 1995, Height and 
Reliability of Edges, ACCV'95 Second Asian Conference 
on Computer Vision, Singapore, II, pp.410-414 
670 
  
  
KEY V 
ABSTE 
A new 
image [ 
image t 
values i 
SW, W 
singula: 
sequen 
points ¢ 
both in 
Trackir 
those « 
Accura 
depend 
order t 
system. 
case Of 
image | 
Efficie! 
parame 
and its 
samplii 
All trac 
this tra 
recogn 
In ord 
period, 
require 
massiv 
In our 
been 
descrif 
dimen: 
and or 
rotatio 
transfc 
rotatio 
been 1 
recogn 
Howe 
is only 
presen 
transfc