Information Technology Reference
In-Depth Information
•
JPEG compression: The distorted images were generated by compressing the
reference images (full color) using JPEG at different bit rates ranging from 0.15
bpp to 3.34 bpp. The implementation was performed by the imwrite.m function
in MATLAB.
•
JPEG2000 compression: The distorted images were generated by compressing
the reference images (full color) using JPEG2000 at different bit rates ranging
from 0.028 bits per pixel (bpp) to 3.15 bpp. Kakadu version 2.2 was used to
generate the JPEG2000 compressed images.
•
White noise: White Gaussian noise of standard deviation σN was added to the
RGB components of the images after scaling the three components between 0
and 1. The same σN was used for the R, G, and B components. The values of
σN used were between 0.012 and 2.0. The distorted components were clipped
between 0 and 1, and then rescaled to the range of [0-255].
Basically, the distortion types and the generation in the stereoscopic image quality
assessment are very similar to those in the 2D image quality assessment. There-
fore, these two image data sets and the corresponding subjective assessment re-
sults can provide a fair comparison of the IQMs between the stereoscopic and 2D
image quality assessments. The performance comparison and analysis will be per-
formed with respect to the LIVE database and the subjective stereoscopic image
quality experiment described in Section 2, respectively.
3.2 Performance Analysis of IQMs on 2D and Stereoscopic
Image Quality Assessment
We performed the 11 IQMs on the 2D and the stereoscopic images, respectively.
As some IQMs use the luminance component only, while others can employ the
color components as well, we transformed all the color images into gray images
firstly, and then computed the image quality using these IQMs. After obtaining the
metric results, a nonlinear regression operation between the metric results (
IQ
) and
the subjective scores (DMOS), as suggested in [42], was performed using the fol-
lowing logistic function:
a
1
DMOS
=
(2)
P
1
+−⋅
exp(
aIQa
(
−
))
2
3
The nonlinear regression function was used to transform the set of metric results
to a set of predicted DMOS values,
DMOS
P
, which were then compared against
the actual subjective scores (DMOS) and result in two evaluation criteria: root
mean square error (RMSE) and Pearson correlation coefficient. The evaluation re-
sults of these eleven IQMs on the quality assessment and the LIVE 2D image data
set are given in Table 2 and 3. According to the evaluation results, some general
conclusions can be drawn as follows.
