We present at figure 10 three linear changes of intensity applied to the
first image Lizard, with respectively
(a) initial image,
(b) =0.4,
=0.3,
=0.3 and
=0.1,
=0.2,
=0.3,
(c)
=0.5,
=0.1,
=0.2 and
=0.2,
=0.3,
=0.0,
(d)
=0.1,
=0.1,
=0.1 and
=0.9,
=0.8 and
=0.7.
The figure 11 shows that it is really necessary to normalize the images before
computing the feature vectors
according to the model described at the equation
4. When the images are not normalized, we obtain approximately 10
of correct matches
whereas the rate increases to 90
when the images are normalized using the method described at
section 2.3. These matches have been obtain using the method of vector comparison
described at section 4.1 and the relaxation technique introduced at section
4.2.
Figure 10: Sequence Lizard with three changes of intensity.
(a) initial image,
(b) =0.4,
=0.3,
=0.3 and
=0.1,
=0.2,
=0.3,
(c)
=0.5,
=0.1,
=0.2 and
=0.2,
=0.3,
=0.0,
(d)
=0.1,
=0.1,
=0.1 and
=0.9,
=0.8 and
=0.7.
Figure 11: Results of matching on the sequence Lizard with three changes of intensity according to our
model. The local normalization of the images has been processed using a window of 21 pixels width.