Prakt. Met. Sonderband 38 (2006) 111
the high _
pre-processing pre-processing no pre-processing Porosity after
procedure with high- | procedure with edge Archimedes
. pass-filter enhancement _
|_100x | 13,340,3% 1 13,4+0,3% ! 13,940,3% _ 11,7%
ı 200x 11,040,6% = 11,7£0,6% _ 11,6+0,6% ‚7%
500x 9,4+1,2% 11,2+1,1% 10,1+1,0% ey, (90
Table 3: Mean values and intervals of confidence (a=0,05; n=20) for porosity after quantification from
images taken at different magnifications, maximised contrast and pre-processed after different routes
From the results presented in Table 3 200x magnification and images without pre-
processing or with edge enhancement were chosen for further examination on the
materials with calculated porosities of 17,5% and 7,3%. From two samples of each
porosity at 200x 10 images each were taken in bright field. The images were pre-
processed and porosity quantified as described above. Quantification without use of the
edge enhancement filter yielded a slightly low result at 7,3% porosity and a slightly high
result at 17,5% porosity. The edge enhancement filter had to be optimized for each
porosity value. By changing the p value in (4) from 0,50 to 0,45 the result for a porosity of
7,3% could be improved from 6,7+0,4% to 7,2+0,3%. At a porosity of 17,5% a p value of
0,55 led to a result of 17,4+0,5% compared to 17,7+0,5% with p=0,50.
Measured porosity: Measured porosity: Measured porosity: Porosity after
no pre-processing edge enhancement edge enhancement Archimedes
ght colours with standard with optimised
value (p=0,50 value
7,0+0,3% i! 6,7+0,4% 7,2+0,3% (p=0,45 7,3%
11,60,6% 11,7+0,6% 11,7£0,6% an Tare
i 18,1+0,5% 17,7£0,5% 17,4+0,5% (p=0,55) 17,5%
Table 4: Mean values and intervals of confidence (0=0,05; n=20) for porosity after quantification from
images taken at 200x magnification, maximised contrast and pre-processed with the noise reduction filter
and the edge enhancement filter (standard and optimised p value)
By utilising equation (5) and the observed values for standard deviation the number of
images necessary for a desired interval of confidence can be estimated. At 200x
magnification, standard deviations in the range of 0,8 to 1,2% were observed. Table 5
shows calculated intervals of confidence (a=0,05) for different standard deviations and
numbers of images. Since the standard deviations calculated from the images are just
estimations of the true standard deviation, the number of images should be chosen rather
higher. For a confidence interval below +0,50%, 20 to 30 images should be used in
quantification.
Estimated values for standard deviation |
Number of images ; 0,80% 1,00% 1,20% |
10 +0,57% +0,72% +0,86%
+0,37% +0,47% +0,56% 1
Fo +0,30% 40,37% +0,45% |
yundary of 40 1 40,26% +0,32% +0,38%
Table 5: Intervals of confidence (a=0.05) for several values of standard deviation and number of images