Information Technology Reference
In-Depth Information
For the distortion caused by JPEG compression, the fluctuation of the perceived
stereoscopic image qualities is more drastic than that of the blurring distortion, as
shown in Fig. 3. The x-axis of Fig. 3 represents the JPEG Q-parameter which de-
termines the quality of the compressed image. In the figure, the quality scores
show an increasing trend as the bit-rates for right view images increasing, and the
maximal DMOS value is below 50. According to the figure, we can find that the
perceived quality of JPEG compressed image is content and texture dependent.
The smoother areas the image contains, the more blocking artifacts are visible. For
example, the
Bowling
image is relatively smoother than other images, as shown in
Fig. 1, and the perceived distortion on this image introduced by blockiness artifact
is, therefore, more visible than on other images, as shown in Fig. 3.
In reference [1], it is stated that for blockiness artifact, the quality of stereo-
scopic images is approximately a mean of qualities of the images presented to the
left and right eyes; while for blurring artifacts, the image quality is less affected by
the poor quality presented to one eye because more weight is given to the input
that has the sharper image, therefore, low-pass filtering (blur) of the images for
one eye is a more effective method for reducing bandwidth than quantization.
However, our experiment indicated that the perceived quality of images with
blockiness artifact is content and texture dependent, and the depth perception de-
grades when the blurring level is increased. Compared with watching the JPEG
compressed images with a similar perceived quality, all participants felt more eye
strain and uncomfortable when viewing the blurred images. Thus, the use of low-
pass filtering instead of quantization for processing stereoscopic images is re-
quired to be explored further.
For white noise distortion, the quality scores show a linear increasing trend as
the noise added to right eye images increasing and the maximal DMOS value is
below 50. According to Fig. 4, we can find that the image content and texture in-
formation have no significant influence on the perceived quality.
100
Art
Bowling1
Computer
Dolls
Drumsitcks
Dwarves
Laundry
M
oebius
Reindeer
Rocks1
90
80
70
60
50
40
30
20
10
0
0
10
20
30
40
50
60
70
80
90
100
Variance of the Noise
Fig. 4
DMOS values of stereoscopic images with white noise
