For these expressions two important pionts have to be mentioned: The AU; coordinate 
depends exclusively on ship speed and is linearily determined from the recording se- 
quence of the image line. using recording time between two positionings. The connec- 
tion to the image is only given, if these positionings are marked in the image (by event 
mark switch). The coordinate input in the equations (8) are YP: and Zpi which are de- 
termined from the image coordinates by using (8) and (7). That Hneans. the processing of 
space coordinates in the coordinate system X; from all the imagery is done before deter- 
mination of (8). Using digital image preprocessing routines, this can be done automati- 
cally and by this all the evaluation process is automized. 
4.2 Determination of Interference Pattern Image Coordinates 
Digitized images are organized in raster coordinates. Each image point is fixed by row 
and column and has an allocated density in integer values between 0 and 255 (B-bit con- 
version). The automatic determination of the image coordinates for the fringe pattern 
is in every single row, which contains the original data. really diffilcult. Figure 11 shows 
why. 
  
ZEILE 100 
  
        
  
  
  
    
   
         
  
  
  
  
ud = 2d 
ZEITMAPK IEPUNG 3 $4 = E. 3 E 
HORAE om Ex d 
28 INTERFEROMETR zn FNAHME DEHSITUMETR. PROFIL ZEILE 199 
  
Fig.11: Interferometric sonar image with densitometric row profile 
While the human eye can detect the pattern by accounting the whole image information. 
it is not possible for a single row. as the densitometric profile shows. Six maxima values 
should be found. These difficulties depend on strong image noise caused by the diffe- 
rent electronic elements for signal processing. Noise has to be removed for further opera- 
tions. 
Tests with simple digital filters like moving average or median filtering gave not the de- 
sired results (Kolouch in SFB. 1982). To find a better working filter function. the image 
was analyzed by Fourier transformation (Castleman, 1979). This could be done one-dimen- 
sionally because the image contains parts, where no sound reflection takes place. In this 
area no image information is received, while the electronic elements are already working. 
so only noise is expected.