Image Processing Reference
In-Depth Information
motion stimuli. This implies that static stimuli induce more visual discomfort when
the relative disparity exceeds a certain value. This value is approximately 1.4
for
planar motion and 2.05
for in-depth motion.
The gradient analysis also reveals that there is no “crossing point” between the
planar motion and the in-depth motion in the positive three-quarters of the disparity
space, i.e., from
0.65 to 1.3
. The in-depth motion stimuli are always more
uncomfortable than the planar motion stimuli in this study. However, for the
condition that the disparity offset is less than 0.65
, we may extrapolate that the
slow in-depth motion stimuli might generate less visual discomfort than the fast
planar motion stimuli. However, further studies are required.
10.3.4 Linear Regression Analysis: Toward an Objective
Visual Discomfort Model
To investigate the influence factors of each motion type, multiple linear regression
analysis is used in this study which attempts to model the relationship between two
or more explanatory variables and a response variable by fitting a linear equation to
observed data.
For the static situation, there is only one possible factor which is the relative
angular disparity. For motion stimuli, the relative disparity offset, disparity ampli-
tude, planar motion velocity, in-depth motion velocity, and their interactions are
possible factors. The stepwise regression function in Matlab was used to select the
most significant factors or remove the least significant factors [
45
]. The output of
the stepwise regression includes the estimates of the coefficients for all potential
factors, with confidence intervals, the statistics for each factor, and for the entire
model. To avoid over-fitting of the model, the Leave-one-out Cross Validation
(LOOCV) method was used for all possible models to find the model with the
minimum averaged RMSE. The selected models are shown in Table
10.4
.
Table 10.4 The linear regression analysis results for all stimuli
Factor analysis
Model analysis
p
-value
(
t
-test)
Factor
Coefficient
D¼
intercept +
Σ
coefficient
factor
Motion type
Planar motion
r
o
1.45
0.0000
Intercept
¼
1.11
R
2
v
p
0.18
0.0000
¼
0.98, RMSE
¼
0.17
10
10
r
o
v
p
0.04
0.0000
F
¼
221.474,
p
-value
¼
4.10
Static stimuli
r
o
2.53
0.0000
Intercept
¼
0.54
R
2
¼
0.99, RMSE
¼
0.15
¼
1176.37,
p
-value
¼
5.45
10
5
F
In-depth motion
r
o
1.23
0.0000
Intercept
¼
2.51
R
2
v
d
0.31
0.0000
¼
0.98, RMSE
¼
0.11
10
9
d
a
v
d
0.45
0.0001
F
¼
147.18,
p
-value
¼
1.8
