Image Processing Reference
In-Depth Information
with the original image
10
to evaluate the quality of reconstruction:
2
. Higher SNR means better reconstruct
20
tion
⁄
performance.
Parameters were define
0.01
,
0.01
,
range with respect to imag
The 3D patch size was 3×
difference between iteration
ence was less than
1
d experimentally. We set
1/3
,
0.02
,
0.1
,
and
were set at 1% inten
ges
L
and
X
, and the maximum iteration number was 2
×3×3 voxels, and search domain was 7×7×7 voxels. T
ns was measured and the program stopped when the dif
5
.
nsity
200.
The
ffer-
esolution image from an original neonatal image. The obser
super-resolution reconstruction and the recovered image wil
tal image for performance evaluation.
Fig. 3.
Simulation of a low-re
image will be the input for s
compared with original neonat
rved
l be
3.3
Results
Experiments were performe
ly. In each experiment, the
their corresponding longitu
include
ed using two imaging modalities, i.e., T1 and T2, respecti
e HR neonatal images were reconstructed with the help
udinal images of same modality. Methods for compari
interpolation (NN), spline interpolation (Spline), non-lo
[6], super-resolution method regularized by low-rank
, and the proposed method. The implementation of NL
te was used
1
. We implemented LRTV by setting
line-based interpolation was used as initialization for NL
ethod.
resentative reconstruction results for T1 (top panel) and
modality, the left panel shows the input neonatal LR im
mage at 2 years of age. The right panel shows the res
p views of selected regions are also shown for better vis
d that the results of NN and spline interpolation meth
acts. NLM results also appear blurry, which is partly
ation is not explicitly considered [6]. LRTV and the p
edge-preserved results. The proposed method achieves
ive-
p of
ison
ocal
and
LM
0
LM,
nearest neighbor i
means upsampling (NLM)
total variation (LRTV) [7],
provided on author's websit
in the proposed method. Sp
LRTV, and the proposed me
Fig. 4 demonstrates repr
(bottom panel). For each m
and its longitudinal HR im
from all methods. Close-up
lization. It can be observed
show severe blurring artifa
cause the blurring degrada
posed method demonstrate
highest SNR.
d
d T2
mage
ults
sua-
hods
be-
pro-
the
1
https://sites.google.com/site/p
pierrickcoupe/
Search WWH ::
Custom Search