Digital Signal Processing Reference
In-Depth Information
Ω
Ω
NP
P
N
NP
P
P
V1
V2 V1
V3V2
V1
Ω
Context-1
Context-2
Context-3
(a)
(b)
(c)
N
V1
P
N
V1
V2
Ω
Ω
Context-4
Context-5
(d)
(e)
FIGURE 10.10
Five context models between two neighboring scales where the coarser scale (top) and finer
scale (bottom) are shown.
{
V
1
,
...
,V
d
}
(
d
=
1
,
2
,
3
)
is the context vector, and
is defined in
Equation 10.31.
The five context models are depicted in Figure 10.10, among which
Context-1 and Context-2 are interscale context models, which are similar to
those used to encourage the formation of large, uniformly classified regions.
7
,
33
Context-5 is an intrascale context model often used in the MRF literature
to ensure local homogeneous labeling with high sensitivity to boundaries.
Context-3 and Context-4 are hybrid inter- and intrascale context models,
which have similar characteristics to those used in References 37 through 39.
We anticipate that those context models have distinct effects on the segmen-
tation results in terms of classification, boundary localization, and boundary
detection. To study their characteristics, we use three numerical criteria to
quantify the segmentation performance. Specifically,
P
a
is the percentage of
pixels that are correctly classified, showing
accuracy
,
P
b
the percentage of
Search WWH ::
Custom Search