Figure 5. Average tracking time per frame in milliseconds on a Dell
Latitude notebook and a Nokia N95 as a function of the number of
shapes tracked in each frame.
Recognition performance is related to the number of shapes in
the shape library and the slant of the viewed shape. To assess this
relation we measured the recognition rate of the system with
different shape library sizes and slants. Figure 6 shows the
recognition rate as a function of the number of shapes in the
library for different slants. The experiment was performed using
the notebook configuration described above, with the camera
fixed approximately 40cm from the shapes. The system was
trained with solid blob shapes from the MPEG-7 shape dataset,
each containing 2 to 8 concavities. For each library size, the
recognition rate was tested on all of the shapes in the library. As
the slant increases, the amount of information in the image