Quality Assessment - 3DTV: Processing and Transmission of 3D Video Signals - page 181

where L refers to the number of pixels of the full image. Structural error of
the depth image/frame is evaluated as the average SC over the macroblocks.

m

1
m

M 3 = SC j (6.23)

i = 1

m

and j refer to the number of macroblocks and its index respectively.

6.4.2.3 Depth Quality Model

Finally, the three components are combined to yield an overall Disparity
Distortion Model (DDM):

DDM = f ( M 1 ( x , y ), M 2 ( x , y ), M 3 ( x , y )) (6.24)

An important point is that the three components are relatively indepen-
dent. For example, the change of relative distance in depth axis and/or
consistency of the perceived depth of the contents in the depth planes will
not affect the structural error of the depth image. Thus the three components
are combined as follows:

M 3 ( x , y )
M 1 ( x , y )

DDM = (6.25)

· M 2 ( x , y ) + k 2

k 2 = 1 is introduced to the denominator to limit the depth quality rating

between 0 and 1, and to avoid instability when M 1 . M 2 is close to zero.
DDM

0 imply the maximum and minimum bounds of
depth quality from the depth image respectively. This is for the reason that,
at maximum depth quality M 1 . M 2 →

= 1and DDM =

0and M 3 → 1 and at minimum depth

quality M 3 → 0.

In practice, one usually requires a single overall measure of the entire
depth sequence. Thus, the mean of DDM index (MDDM) is evaluated to
predict the depth quality from the depth sequence:

M

1
M

MDDM ( X , Y ) = DDM ( x i , y i ) (6.26)

j = 1

where X and Y are the reference and the distorted disparity signals respec-
tively, x i and y i are the contents at the j th

frame, and M is the number of

frames in the depth map.
Thus, as shown in Equation (6.9) the overall depth quality from 3D video
(i.e. effects from both colour texture video and depth map) can be modelled
as follows:

VQM ] α · MDDM β

Depth _ quality = [1 − (6.27)