Biomedical Engineering Reference
In-Depth Information
2.4
Quality Metrics
For the quantitative evaluation of filters' performance, two quality metrics used in
[
23
,
26
]and [
28
] were applied. The metrics used are the mean squared error (MSE)
and the edge preservation parameter ().
The MSE is given by:
X
L
X
J
I.l; j/
I
f
.l; j /
2
;
1
LJ
MSE
D
(13)
j
D
1
lD1
where I.l; j/ denotes the original image, I
f
.l; j / denotes the filtered image,
L and J are the number of pixels in row and column directions, respectively.
To evaluate the edge preservation we use the correlation coefficient (
14
)asin
[
26
,
28
], where I and I
f
are the laplacian of the original and filtered image,
respectively, and the I and
I
f
the respective mean values.
P
.l;j /
I
I
I
f
I
f
s
X
D
.
(14)
I
X
.l;j /
I
I
I
I
f
I
f
I
f
I
f
.l;j /
In order to quantify the smoothness of homogeneous regions (H ) the equivalent
number of looks (ENL) is computed according to (
15
).
X
H
h
h
1
H
D
ENL
,
(15)
h
D
1
where
h
and
h
are the mean and variance, respectively, of the regions H .
3R su s
In order to evaluate the performance of the filter herewith presented, eyes of
30 healthy volunteers and eyes of patients with age-related macular degeneration
(20), diabetic retinopathy (23), cystoid macular edema (2) and choroidal neo-
vascularization (13) underwent the high-definition spectral domain Cirrus OCT
using both the 200
200
1;024 and the 512
128
1;024 Macular Cube
protocols.
The volumetric scans were filtered using the here proposed 3D nonlinear com-
plex diffusion filter (3D-NCDF), as well as using its 2D version [
28
], the Perona-
Malik filter (PM) with the diffusion coefficient d.
=k/
2
/,where
jr
I
j
/
D
exp.
.
jr
I
j
k
D
50 [
24
], and the adaptive Lee filter with a window of 3
3 pixels [
22
].
