3. Digital Image Processing 
  
With the block diagram of figure 1 the main parts of an image 
processing system have been described in principle. Some of these 
parts have been selected and will be discussed in detail within 
this chapter. 
3.1 Image enhancement 
Image enhancement is a part of image preprocessing and usually trans- 
forms a given image into another twodimensional representation 
being more suitable either for machine or human interpretation. 
3.1.1 Pixel oriented operations 
  
In pixel oriented methods the grey value B of the output 
: 2'ij 
image only depends on the corresponding grey value B, ij of 
the input image (equ. 6). 
f (B (6) 
1,ij) 
The function f may depend on i and j or on the input grey level 
B,. In these cases the operation is called inhomogeneous, whereas 
in the case of homogeneous operations the function f is indepen- 
dent of the location and the information of the input element. 
In figure 8 several examples for possible density manipulations 
are shown. Figure 8a represents a linear, homogeneous operation 
where the function f is only a multiplicative constant value. 
Quadratic, exponential or threshold operations are illustrated 
in the examples 8b - 8d, representing nonlinear functions. A lo- 
garithmic function e.g. may be applied in order to simulate the 
logarithmic evaluation of the human eye. 
- 10 - 
Q 
“ 
In many applications the grey level resolution of the original 
image has to be reduced either for further machine evaluation 
(e.g. multispectral feature extraction) or for human interpre- 
tation of an image using a display with low grey level resolu- 
tion. In these cases a technique called "histogram equalization" 
is used before reducing the grey level resolution. Applying this 
technique to an image the grey levels with a high frequency of 
occurence are distributed into a large number of grey levels 
whereas grey levels with a low frequency of occurence are gathered 
into a few number of grey levels. Figure 9 shows in principle the 
frequency occurence of the grey levels in an image before and af- 
ter the histogram equalization, whereas figure 10 demonstrates 
the performance of the technique when being applied to real images. 
Another method of those density manipulations has to be used in 
change detection where the changes in two similar images have 
to be recorded. Figure 11 shows diagrams of the corresponding grey 
levels in two images B, and B». Using two images which are totaliy 
equal, the corresponding points are located on the dotted line 1. 
According to changes in illumination or according to different 
photographic materials, in real images the corresponding pixels 
form a region, illustrated by region 1 in figure 11a where the 
pixels are arranged on both sides of the dotted line 2. Therefore 
the grey levels of image B, have to be adapted to the grey levels 
of image B, in such a way that the grey level distribution in both 
1 
images becomes similar. This results in a diagram according to 
figure 11b where the two peaks 1 and 2 represent the real changes 
in the two images. 
The human eye has a better resolution for different colours than 
for different grey levels. Therefore in many interactive systems 
a pseudo-colour representation (/19 - 21/) is used to display 
grey value images to a human interpreter. By this technique each 
grey level of the original image is represented by a special 
combination of the three primary colours red, green and blue.