Digital Signal Processing Reference
In-Depth Information
Where,
is the ith foreground sample and
is the jth background sample.
i
j
F
B
is the distance between the foreground/background pair:
R
(
F
i
,
B
j
)
d
ˆ
i
ˆ
j
I
−
(
α
F
+
(
−
α
)
B
)
i
j
R
(
F
,
B
)
=
d
F
i
−
B
j
Where,
αˆ
is the estimated alpha value defined by:
j
i
j
(
I
−
B
)(
F
−
B
)
ˆ
α
=
2
F
i
−
B
j
And,
and
are two weights defined as:
{
i
j
ω
(
F
)
ω
(
B
)
}
2
ω
(
F
i
)
=
exp
−
F
i
−
I
/
D
2
F
And
{
}
2
ω
(
B
j
)
=
exp
−
B
j
−
I
/
D
2
B
On the other hand, the affinity between two antibodies is defined as:
(3)
Hypermutation: antibody's hypermutation operation in IFEN is to bring change to the
antibody network, so we design the hypermutation used in matting as:
aff
=
(
F
i
,
B
j
)
−
(
F
u
,
B
v
)
+
(
i
−
u
)
2
+
(
j
−
v
)
2
bb
k
(
i
,
j
)
=
(
i
,
j
)
+
int(
ωβ
R
)
(4)
k
k
k
ji
refer to the current indexes of fore-
ground/background pair and the indexes in the k-th hypermutation, respectively.
k
β
=
0
.
5
(
i
,
j
)
(
,
)
Where
,
and
k
R
is the maximum indexes of
foreground and background. More details of this procedure refer to [10].
is a random number with normal distribution.
ω
Fig. 3.
The flowchart of IFEN
