Graphics Reference
In-Depth Information
Figure
.
.
Binary images: artificial image containing four circles with different gray values (let),
binary image generated by Bernoulli experiments with pointwise probabilities proportional to the gray
values in the let image (center) and reconstructed image of pointwise probabilities (right)
and to replace the estimate θ
(
k
−)
j
in (
.
,
.
) by θ
(
k
−)
ij
(
k
−)
i
θ
(
k
−)
j
=(
−
.
N
)
+
θ
(
k
−)
i
for all subsequent iteration steps.
We use a simple artificial example to demonstrate the performance of the pro-
cedure. We start with the image displayed on the let of Fig.
.
. he image of size
(
k
−)
i
.
N
is composed of four regions with distinct gray values. he central image is
generated by pixelwise Bernoulli experiments with probabilities of
.
,
.
,
.
and
.
,respectively, forthe fourregions. he imageon the right of Fig.
.
provides the
reconstruction obtained by AWS using a maximal bandwidth h
max
=
. he value
of λ selected by our propagation condition is λ
.
.
he noisy image on the let of Fig.
.
is constructed using the same image struc-
ture.Noweach grayvalue isaPoisson count ofintensity
.
,
.
,
and
.
,depending
on the location of the pixel within the image. he right image again provides the re-
construction. A maximal bandwidth h
max
=
and the value λ
.
provided by
=
=
the propagation condition (
.
) for
-D Poisson images are used.
Figure
.
.
Poisson images: image generated by Poisson experiments with pointwise intensities
proportional to the gray values in the let image of Fig.
.
(let) and reconstructed image of pointwise
intensities (right)
