shows less details and so on. The subsequent PC score images show
great amounts of noise or random variations in intensity due to the
fact that the major features of the data have already been captured in
the first few PCs. Therefore, the first three PC score images were then
combined altogether to form a pseudo-color image (one score image
stands for one color channel). It is obvious that fat portions are
explicitly distinguished from the loin eye and the other adjacent
muscles for all pork classes (Fig. 7e). However, score images had
subtle differences in color among the pork classes, thus making
difficult to identify each class. It was necessary to apply a mapping
transformation to enhance these differences, allowing the analyst to
clearly distinguish between different classes of pork. Therefore, the
constructed pseudo-color image was converted to an indexed image
by using a color map of only limited number of colors instead of using
the full color range. The resulting indexed image is called classification
image in this context because it shows the result of classification that
could be presented in examined sample (Fig. 7f). The indexed image
uses a directmapping of pixel values to the values existing in the color
map. The color of each image pixel in the indexed image is determined
by using the corresponding value in the color map. Thus, pixels in the
composed pseudo-color image having same features tend to be
indexed with the same values in the color map. As a result of
conversion to indexed image, PSE class is presented in green color,
RFN in red, DFD in blue and fat in yellow. Fig. 8.A2, B2 and C2 showthe
indexed images for the same samples obtained for the loin eye alone.
Accordingly, sample 8A was classified as PSE, sample 8B as RFN, and
sample 8C as DFD. Detailed observation of these samples show that
there are regions of different classes in certain parts of the loin eye,
but these regions are small compared to the larger regions which