Image Processing Reference
In-Depth Information
between the proposed
Near NR
quality metric versus average PSNR (dB) of
rendered left and right view quality for color and depth map sequences respectively.
These figures consider sequences encoded at both
QP
30. These
scatter plots also demonstrate a higher degree of correlation between the proposed
quality evaluation method and the rendered quality of left and right views with
DIBR. These findings suggest that the proposed method can be used to measure the
quality of individual color and depth map views or approximation of true 3D
perception (MOS) on-the-fly with a minimum overhead for side-information.
¼
10 and
QP
¼
Conclusion
This chapter proposed a
NR
quality metric based on edge detection for color
plus depth 3D video compression and transmission. Since the edges of the
color image and the corresponding depth maps represent different depth
levels and the basic structure of the color and depth map objects, edge
information extracted at the receiver side is used as a measure to quantify
the structural degradation of these sequences. Since only statistics related to
luminance
and
contrast
information of the original image sequences are
required for image quality evaluation as side-information, the proposed
method is close to a complete
NR
quality metric, hence referred to as
Near
NR
in this chapter. Results show a good approximation of the associated
FR
quality metric for all considered
PLR
s, compression levels, and wireless
impairments. However, due to the abstract level of information used to
measure the structural degradation with the proposed method, results may
differ from the
FR
method. In order to obtain matched results, the proposed
method can be calibrated against the
FR SSIM
metric based on experimental
results. In this case the performance is almost indistinguishable from the
FR
method, with an R-square value of 0.94 and 0.98 for color and depth map
sequences respectively.
This suggests that due to the practical problems associated with the usage
of
FR
methods for online system optimization,
NR
quality metrics as
described in this chapter are an acceptable compromise for the design of
QoE-aware 3D multimedia systems.
References
1. Le-Callet P, Moeller S, Perkis A (2012) Qualinet white paper on definitions of quality of
experience, Version 1.1. European network on quality of experience in multimedia systems
and services, COST Action IC 1003, June, 2012
2. Martini MG, Mazzotti M, Lamy-Bergot C, Huusko J, Amon P (2007) Content adaptive
network aware joint optimization of wireless video transmission. IEEE Commun Mag 45
(1):84-90
3. Meesters LMJ, IJsselsteijn WA, Seuntiens PJH (2004) A survey of perceptual evaluations and
requirements of three-dimensional TV. IEEE Trans Circuits Syst Video Technol 14(3):381-
391