Image Processing Reference
In-Depth Information
Fig. 9.5 Measured image quality of
Ballet
sequence at
QP
¼30 using
FR
and proposed
Near NR
method. (a) color image quality, (b) depth map quality
9.3.1 SSIM Based Near NR Quality Evaluation (“E-SSIM
C
”
and “E-SSIM
D
” Methods)
Figures
9.6a-e
show the scatter plots for the measured image quality (i.e., quality of
both color image and depth map) using the
FR
and the proposed
Near NR
methods for
the
Orbi
,
Interview
,
Ballet
,
Breakdance sequences
, and all the sequences in general
respectively. Each point of this plot corresponds to the measured quality of an image
frame using both
FR
and proposed method for a given compression level (QPs ranging
from 1 to 50),
PLR
(0, 3, 5, 10, and 20 % PLRs), and
E
S
/
N
0
. According to these figures
it is evident that there is a close relationship between the quality ratings of these
methods regardless of the compression and PLR/(
E
S
/
N
0
) level being considered. In
order to analyze the correlation between the reference and the proposed methods, the
quality ratings are approximated by a second order polynomial [see Eqs. (
9.14
)and
(
9.15
)] for all the sequences at different compression levels and packet loss rates.
Fitting curves for the proposed and
FR
methods are shown in each of the scatter plots
in Fig.
9.6
to show the quality of the fitting. This shows a higher degree of correlation
between the proposed method and
FR
method for both color and depth map images.
2
SSIM
C
p
1
E-SSIM
c
ð
Þ
þ
p
2
E-SSIM
c
ð
Þ þ
p
3
ð
9
:
14
Þ