The box counting procedure (see [4][3]) is suitable for binary images. The plain Box 32 button calls the box count algorithm with boxes at a maximum size of 32 pixels. The result for the image in fig. 9 is shown in fig. 27.
Figure 27: Box count 32 result
The natural logarithm of the box size is plotted on the horizontal axis and the logarithmic total area of boxes containing at least one white pixel is plotted on the vertical axis.
It is possible to exclude or include certain box sizes or regions of sizes
with the mouse. In fig. 27 only the region between the two
vertical lines was taken for the calculation. The smallest box size is
and largest box size is
.
The box dimension is determined from the slope of the regression line.
With the formulas and the dimension specification given in [4]
the dimension is .
The boxcount 64 algorithm, that is a maximum box size of 64 pixels is used,
yields according to fig. 28 ,
which matches nicely with the box 32 dimension.
Figure 28: Box count 64 result
It is demonstrated in fig. 28 that the coordinates of the markers are presented if the mouse pointer is placed on the mark.
Fig. 29 shows the result of the analysis with a maximum box size
of 128 pixels with the box sizes 1 and 2 excluded.
The dimension is ,
Figure 29: Box count 128 result
The results of the boxcount 32 or boxcount 64 procedure and the boxcount 128 procedure with boxes of size 64 and 128 excluded might be slightly different because the image is resized by a multiple of 32, 64 or 128, respectively.