Prakt. Met. Sonderband 38 (2006) 109
>ssed and TIFF to 8Bit greyscale TIFF-format and the contrast maximised by histogram stretching.
rials were The obtained pictures were quantified by setting the threshold automatically and
comparing the number of pixels with grey-values equal or lower than the threshold with the
total number of pixels. Porosity was calculated after
Macro-
hardness N
HV-30 V,=A4, == (3)
87 Na
105 |
138 With v, the volume fraction of porosity, A, the area fraction of porosity, N, the number of
en pixels associated with porosity and N, the total number of pixels [4]. Two different routes
of digital image preparation were examined afterwards. Since the aim of the image
preparation was to obtain well defined boundaries between pores and matrix, the images
were treated with a high pass or sharpness filter to increase the gradient in colour between
pore and matrix. The used filter operates with a 3x3 environment, the central pixel has a
ed after a weight of 9, the surrounding pixels have weights of -1 [5]. After applying the high pass
sur steps. filter to the images, the boundaries of pores and matrix showed a steeper gradient in grey
ed in an values as visible by comparing the digitally zoomed sections of Fig. 1 and Fig 2. As an
» following alternative route a noise reduction filter (sigma filter) and an edge enhancement filter were
with 3 pm used. The edge enhancement filter operates in a 5x5 environment, the grey value of the
ary for the central pixel is compared with the grey values of the surrounding pixels after
leps. After
examined Greentral 2 Gor =n: LET Gin) (4)
ler and or
of pores. with G.,,,., the grey value of the central pixel, G,,, and G,, the maximal and minimal grey
d out and value observed in the environment and p a parameter with a value between 0 and 1. If
> Se the relation (4) is true, G.,,,, is substituted by G,,, , otherwise G.,,,,, is substituted by G,,, .
methanol Use of the edge enhancement filter with p=0,5 resulted in well defined boundaries as can
be seen by comparing the digitally zoomed portions of Fig. 1 and Fig. 3.
yolishing 2
sli 2.4 RESULTS OF POROSITY QUANTIFICATION
im
50 At each magnification the effects of the pre-processing methods described above on
_ —_ porosity quantification were evaluated [6]. The 20 images at each magnification taken of
1in the material with a calculated porosity of 11,7% were pre-processed after both methods
—— and the results of quantification for porosity compared with quantification without pre-
N processing and the calculated porosity. From standard deviation and the number of
C pearlitic images intervals of confidence cI were calculated after
Ss
Cl=t,..1 Te (5)
with £ the value of the Student statistics for n-1 degrees of freedom and a level of
steels by significance 0<a <1, n the number of images used for calculating the mean value and the
ation were standard deviation. The best results were obtained at 200x magnification. Quantification
e material with 100x magnification yielded too high results for all pre-processing methods examined,
procedure quantification with 500x magnification yielded too low results in combination with rather
OLD broad intervals of confidence. At 200x magnification the best results were obtained without
i