problems: a shape of contour line is often much 
different from the neighbor contour line; and there 
are very few information of contour in low land 
area (Figure 1). 
By the way, DEM can be divided into two 
models: Raster type; and Vector type. A 
represented vector type DEM is TIN (Triangulated 
Irregular Network) model. This model is supported 
by almost GIS software. An accuracy of this model 
depends on network condition. The triangle 
sometime rejects detail of complicated contour line. 
So, an accuracy of raster type DEM is higher than 
TIN. Therefore, an objective DEM in this study is 
selected raster type. 
Generally, an interpolation of contour line is 
calculated pixel by pixel for raster type DEM. A 
profile or a window is usually used for the 
interpolation. However, when we use a small scale 
contour map, generated DEM from those methods 
have not enough accuracy. So, we developed new 
method which is based on intermediate contour 
line derivation. And an efficiency of the 
developed method was discussed by comparing 
with existing methods. 
For evaluation, comparison item is usually 
used only elevation. However, DEM is applied to 
many analysis such as hill shading, run off 
analysis, slope stability analysis and so on. Those 
applications require a total balance which is 
combination of elevation, inclination and slope 
aspect. Therefore we tried to evaluate on many 
items. For example elevation, inclination, slope 
aspect, undulation, stream pattern and slope 
stability was used. Finally, a relationship 
between contour line interval and accuracy was 
concluded. This information will be very important 
to generate DEM from a contour map. 
2. EXISTING INTERPOLATIONS 
The most popular method of the 
interpolation is using a profile which include 
target pixel. In this paper, this method is called 
"profile method". An elevation of the target pixel 
can be estimated by curve fitting from crossing 
points of contour line along the profile. Linear 
equation, polynomial equation, second order 
equation or Spline function is used as curve fitting 
equation. Figure 2 shows a shaded image of 
generated DEM by Spline function from a test 
contour line map (Figure 1). The contour line 
interval was spread purposely in order to see 
reliability of the result. In this image, there are 
profile. Searching the most suitable direction of 
profile is difficult in this method. 
Another popular method of the 
interpolation is using a window which a target 
pixel is located on the center. In this paper, this 
method is called "window method". Consisted 
pixels of contour lines in the window are used as 
random points. An elevation of the target pixel can 
be estimated by weighted mean calculation. Figure 
3 shows a shaded image of generated DEM from 
  
many radiated noises which are along calculated Figure 3 Shaded DEM from Window method 
848 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B4. Vienna 1996