106 Prakt. Met. Sonderband 38 (2006)
2. QUANTIFICATION OF POROSITY BY IMAGE ANALYSIS dur
pol
Porosity was to be quantified from images taken in bright field, where matrix and pores afte
show significantly different intensities. While the polished matrix of the materials exhibits a ste
narrow range of nearly white colours in greyscale pictures, pores are significantly darker in MIC
colour, mainly black and dark greys. By defining the threshold in greyscale-space between Sal
colours representing the matrix phase and colours representing pore phase the amount of intr
porosity in an image can easily be obtained by counting the pixels representing the Ste
phases. The main source of error in this procedure is the gradual change from dark to Se
bright colours in the transition from pore to matrix. Since the result of the quantification po!
step is influenced significantly by the definition of the threshold, a reliable method for ne
porosity quantification of sintered steels has to define the boundaries between pores and pic
matrix in a way that correct results can be obtained. To evaluate the treatment of the de
digital images prior to the quantification step regarding its usability for porosity cal
quantification, materials with known porosity (calculated from theoretical and sintered CO!
density) were used as testing materials. 2
sm
qui
to |
but
2.1
A!
wa
9,8
50
10
Aft
11
en:
Co!
Arc
Fig. 1: 200x micrograph (8bit greyscale with maximised contrast) of a pressed and sintered steel with to :
calculated porosity of 11,7%: Gradual change from dark to bright at the boundary of pore-phase and matrix- inf
phase. Main image 2576:1932 pixel (627,1um:470,4um); upper right corner digitally zoomed 170:128 pixel ca
(41.4um:31.2um); lower right corner digitally zoomed 44:33 pixel (10,7um:8.0um)
Porosity quantification by image analysis is afflicted by both systematic and statistical
errors. The main sources of systematic errors are sample and image preparation [2]. A
correct sample preparation in case of opened pores could be ascertained by comparing Wi
the quantified porosity of the samples with the known porosity, if a way of image
preparation without error was available. The quality of image preparation could be checked or
by comparing the results of porosity quantification for a correct prepared material with the ce
known porosity. Since sample and image preparation should be controlled against each
other, the intrinsic error can not be quantified. A suitable way to minimise these sources of
error is careful visual evaluation of the samples between polishing steps. The aim of
sample preparation for porosity quantification is the removal of the material layer deformed