An appropriate metallographic procedure was performed on both samples, in order to eliminate the 
scratches and artifacts, to assure a good contrast between matrix and boundary. In automatic image 
analyses, this type of defect can lead to an erroneous result, because the equipment works with gray 
levels for the detection of the boundaries in the matrix. The metallographic etchings used were Nital 
2% for the low alloy steel and acetic-glyceregia for the austenitic stainless steel. 
The grain size measurements were carried out in an analyser image system with software based on 
ASTM E 112 and ASTM E 1382 standards. In the low alloy steel sample, with the application of the 
Histogram Equalization, Gray Invert , Sharpen, Threshold , Filling, Erosion and Thicken filters , as 
well as logical operations and automatic edition of bitplanes, were possible to enhance the contrast 
between grains, allowing a more accurate identification of the boundaries , and also eliminating the 
artifacts presented inside the grain that could be accounted as a new boundary. For the austenitic 
stainless steel sample, the same filters and logical operations were applied , however the verification of 
the particles aspect ratio for twin detection was also necessary in order to include them in their original 
grain, assuring that they could not be analyzed as a new grain ( 3-5). 
3. RESULTS AND DISCUSSION 
The Figure 1 presents micrographies of the low alloy steel sample, before and after the application of 
the algorithm containing filters and logical operations, where the artifacts inside the grains (1), and 
grain boundaries discontinuities (2) are evident, as shown in fig. 1A. After image treatment ( fig. 1B), 
we can notice that the equipment could detect new existing grains, which means a decrease in the grain 
size and consequently a change of the ASTM value that vary inversely proportional to the grain size, 
from 6.37 to 8.43. 
In the figure 2 are presented micrographies of the austenitic stainless steel, before ( fig. 2A) and after 
(fig. 2B) the image treatment, as well as two illustrations, one considering the twin boundaries as an 
effective boundary (fig. 2C) and the other one considering the twins inside their original grain (fig. 
2D). In this case, the exclusion of the twin boundaries during grain size measurements lead to an 
increase of the grain size. 
The results of the grain size measurements obtained for both samples, before and after the image 
treatment are presented in Table 1. For the low alloy steel sample, the result after image treatment for 
the ASTM Grain Size presented 32.34% of difference when compared to the measurements made 
before treatment(8.43 and 6.37). In the austenitic stainless steel sample, the result after image 
treatment presented 41.43% of difference (6.30 and 8.91). In both cases, the values obtained after the 
image treatment are very close to their respective true values, 8.50 to the low alloy steel and 6.00 to 
the austenitic stainless steel, obtained by interlaboratorial comparison. 
3. CONCLUSION 
The results obtained on both samples showed a significantly change in the ASTM Grain Size, which 
means that the automatic image treatment, with the application of filters and complementary logical 
operations. is very effective in minimizing error in the grain size determination. 
176