images
size and number of lines, we use Pearson’s correlation
tests. Obtained results show that there is a moderate relationship
between execution time and image’s pixel size (corr. =
0.535). Meanwhile, execution time has a relatively high
correlation, at 0.85, with the number of lines. Figure 5 indicates
the relationship along with its linear model.
As the above discussion attests, processing on architectural
diagrams, maps and blueprints might take a lot of time
and system resource. Among the 4 most time-consuming
images showed in the Figure 6, only the 2nd one is a UML
CD. Therefore, an early recognition of such the images will
significantly decrease the execution time. It can be done by
applying template-based matching and image pyramids [27].