cures is 
erometric 
arrow 
Fig. 7: Filter Design and Result (90 ©) 
-sufficient for following automatic interferometric image coordinate 
detection. 
Therefore filters should consider information in a wider angle area, 
and in addition a low pass filter should be applied to suppress high 
frequency noise too. In the next examples the filter angle is taken 
between + 60 9 in the frequency domain and additional low pass filter 
with different degrees are superimposed (figure 8 - 10). 
The effects of filtering for the specialaim of maximum detection in 
ISSS-images can be seen more dearly in one-dimentional image profiles 
(figure 11) corresponding to figures 8 - 10. 
The best result of all tests is obtained by filter 'C'. It is a good 
compromise between noise reduction by low pass filtering and fringe 
enhancement by direction filtering. The filtered image 'C' is now taken 
for further processing. It was found that now with an additional one- 
dimen ional 'moving median', which works row by row (result see figure 
12), simple line-following procedures can be applied for the desired 
maximum detection. Figure 13 a shows a binary overlay of the automatic 
detected maximum coordinates from the filtered interferometric image 
with the original image. To optimize these results following image 
processing techniques (e.g. enveloping) can be obtained, which work 
directly on the artifical binary coordinate image. This is - again as 
overlay - presented in figure 13b.