Quality Assessment - 3DTV: Processing and Transmission of 3D Video Signals - page 173

the visual acuity approach (Measure 3). Specifically the scores obtained for
the left and right image have been weighted by using the visual acuity of the
observers. The rationale behind this approach is that the perceived quality
could decrease along with the visual acuity of the observer. The maximum
visual acuity of the observer is defined to be equal to 1. A Snellen eye chart
is used to measure the visual acuity of the test subjects. In Measure 4, the
objective quality rating of colour texture video is considered.
Some common objective metrics have been compared to the actual MOS
results obtained from subjective experiments. The correlation of the results
has been assessed using the procedures described in Section 6.3.1. According
to Table 6.3 all quality models are generally acceptable in predicting the image
quality attribute of 3D video. However, the results show that Measure 4 of
VQM has the best correlation with respect to subjective ratings for predicting
image quality. Thus, the VQM quality rating of the colour texture video can
be effectively used to predict the image quality of colour-plus-depth-based
3D videos.

6.4.2 ModellingtheDepthQualityof3DVideo

From tests 1 and 2, it is noted that depth quality is affected by both colour tex-
ture video (contributing to monocular cues) and the depth map (contributing
to binocular cues). However, experiments with so-called random-dot stere-
ograms show that binocular and monocular depth cues are independently
perceived. In a similar way, human brains will independently perceive
monocular and binocular degradation in a 3D video [7, 18]. Furthermore,
the added value of strength of depth by the depth map (3D viewing com-
pared to 2D) can be identified as the effect from the superimposition of
binocular cues on monocular cues. Thus, the overall depth quality can be
modelled as follows:

Depth _ quality = f ( D M , D B ) (6.8)

D α M D B Depth _ quality = (6.9)

where D tot refers to the objectively evaluated inclusive depth quality and
D M and D B refer to the contribution to depth quality from monocular and
binocular cues respectively. are positive constants that define the

relative importance for depth quality. Due to the independent nature of
monocular and binocular cues of depth perception, first the contribution
from the monocular cues to overall depth quality is evaluated and, second,
the added value from the binocular cues is modelled.

α and β

6.4.2.1 Contribution from the Colour Texture Video

If no depth map is presented, similar to 2D video, depth quality is judged only
from the monocular cues extracted from the colour texture video. A fourth