ilter size in this 
filter because of 
However, since 
js in the edged 
difficult to judge 
shape of pattern 
ie edged image 
  
  
VR i M "y 4 A EA a 
Z Ro E ^ 
d \ ! À S Cr Tl NPY 45, : 
A x i vi n i Aet A y hs i M. J ; i" 4 
+ ~ SES : n Sohnes es e ÿ » eb M BI 
su: Y H “el x Rem E. i ^ y. $ Xr LN AR 
  
  
The archaeologists are collecting the scientific data 
including the supposition of years and classification of 
cultural systems by checking to see a general shape of the 
artifact, the shape of pattern and the directionality of 
pattern from the ortho projection image and drawn (edged) 
image which is made of the image and they have formed a 
conclusion saying that an extra information of a noise 
included in the image is unnecessary. 
Thus, in the present situation, the edged image obtained 
by using the filtering technique is handled by 
characterizing it generally as a thing used as a reference 
given immediately after photographing or for drawing by a 
manpower. 
Therefore, in this investigation, it has been decided to 
employ an edge analysis technique using a reliability for 
making an edged image having even higher definition 
than the result obtained by using the conventional LG filter. 
3.2 EDGE METHOD USING RELIABILITY 
The edge detection, broadly classified, is composed of 3 
processes 
(1) detection of features of edge from image, (2) judgment 
of whether extracted features are on edge and (3) 
connection of edged points, and the detection of the 
features of the edge is greatly significant in this work. 
Mr. Sugiyama et al. have proposed a method for extracting 
the features by setting 2 threshold limit values relative to 2 
components called the height and reliability of edge and 
the elements obtained from these components 
(T.SUGIYAMA et al., 1995) . 
This height of edge is a variation obtained from a relation 
of brightness with the periphery at a point of attention and 
the reliability is an index representing the scale of 
influence of noise . 
The height of edge(edge variation, h (x,y)) can be 
obtained by applying an expression(1) shown below to an 
image becoming an object, which is representing the 
variation of edge at the periphery of the point of attention. 
Moreover, o in the expression(1) denotes dispersion of a 
weight function (w(r)) and it is obtained from a relation with 
a size (WS) of window (calculation region) set at a time of 
calculation. And, it has been set as WS = 20. 
  
h(x,y) = JH (XY) + Hy (XY) — «0m 
Where; 
h, (x,y) = 
EL. f(x +rcos0,y +r sin0)w(r) cos0drdo 
h, (X, y) = 
i L f(x +r cos6, y + r sin6)w(r) sin6drdo 
667 
  
  
f(x,y) : Brightness value at (x, y) 
At the same time, the reliability (degree of noise influence, 
r(Xy)) can be obtained from an expression(2) shown 
below and it is handled as the scale of noise influence and 
obtained at a range of 0 to 1. 
r(x,y) = hoy). -- (2) 
20,(X.y) 
Where; 
2 
o2(x,y)- [i f(x  rcosó,y « rsinó) w(r)drdo 
-( [ f f(x rcosó y « rsinó) w(r)drd6)? 
For detecting the features of edge, it makes it a rule to 
judge whether the edge height (h (x,y)) at each point fulfills 
an expression(3) (shown below) represented by using the 
above expressions(1) and (2) and preserve only an edge 
height fulfilling the condition as the edge. As a threshold 
(ht") in this case, the following two ones have been set and 
a threshold corresponding to either condition is preserved 
as the edge and a threshold not corresponding is deleted. 
Moreover, the threshold for the height of edge and 
reliability set in this case is set manually by the operator 
based on the result of edged image which is an output 
result. 
1 
hi < hoyos yep (3) 
20 
Conditions : 
(1) Edge detected by high threshold (htu) 
(2) Edge detected by low threshold (htl) and having 
reliability(RelyLim) which is higher than threshold 
4. EDGED RESULT 
Figure 3 shows a height images, Figure 4, an image of 
reliability and Figure 5 shows edged images, 
respectively. The height images are obtained by imaging 
the results brought by applying the expression(h(x,y)) used 
for obtaining, the height of edge shown in the foregoing 
expression(1) to the image given in Figure 2. 
Consequently, an image in which a pattern of such an 
image including noise as that made by the conventional 
filtering technique has been acquired. The image of 
reliability is obtained by imaging the result brought by 
increasing the result obtained by the expression (r(x,y))