Face Detection and Modeling for Recognition - Biometrics Research ...
Face Detection and Modeling for Recognition - Biometrics Research ...
Face Detection and Modeling for Recognition - Biometrics Research ...
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
[ (<br />
∂Φ<br />
∂t =∇Φ µ 1 div(g ∇Φ<br />
)<br />
|∇Φ| ) − µ 2r − αg<br />
− ∑ |I − c i | 2 + ∑ ]<br />
|I − c j | 2 , (5.16)<br />
i<br />
j<br />
g 0<br />
g =(1 −<br />
max(g 0 ) )2 , g 0 = log(1 + |∇I| 2 ) 2 (5.17)<br />
{<br />
1/dt dt ≠ 0<br />
r =<br />
(5.18)<br />
MAXDIST dt = 0<br />
where µ 1 is a constant, µ 2 <strong>and</strong> α are constants in the interval between 0 <strong>and</strong> 1, I<br />
is the image color component, c i <strong>and</strong> c j are the average color components of facial<br />
component i over region Ω + i<br />
<strong>and</strong> component j over Ω − j , respectively, r is the component<br />
repulsion, dt is the absolute Euclidean distance map of the face graph, <strong>and</strong><br />
MAXDIST is the maximum distance in the image. We further preserve facial topology<br />
using topological numbers <strong>and</strong> the narrow b<strong>and</strong> implementation of level-set functions<br />
[183]. The preliminary results are shown in Fig. 5.13 with evolution details <strong>and</strong> in<br />
Fig. 5.14 without the evolution details.<br />
The facial distinctiveness of individuals can be seen from the changes among the<br />
generic, the fine fitted, <strong>and</strong> fine de<strong>for</strong>med face templates, shown in Figs. 5.13(a),<br />
5.13(d), <strong>and</strong> 5.13(e).<br />
Comparing the two approaches <strong>for</strong> de<strong>for</strong>ming interacting<br />
snakes, we believe that the first approach, parametric active contours, is better suited<br />
to the de<strong>for</strong>mation of semantic face graphs.<br />
129