Information Technology Reference
In-Depth Information
These distortions reflected a broad range of types of image impairment, from
smoothing to structured distortion, image-dependent distortions, and random
noise. The distortion levels covered a wide range of quality degradation from im-
perceptible levels to high levels of impairment.
A subjective human trial, based on the ITU-R recommendation BT.500 [30], as-
sessed the quality of the generated distorted stereoscopic images against the original
images. A double-stimulus continuous quality-scale (DSCQS) method for stereoscopic
image assessment was employed. The DSCQS method was cyclic, in which subjects
viewed a pair of pictures with same content, i.e. a distorted image and an original im-
age, and were asked to assess the qualities of both two images. The subjects were pre-
sented with a series of stereoscopic image pairs in a random order.
In the experiment, polarized glasses were worn in order to separate the left and
right images on a single screen to the appropriate eyes. The experiment was con-
ducted in a dark room, with constant minimal light levels. Twenty non-expert sub-
jects participated in the quality evaluation, whose ages vary from 20 to 25 with a
mean of 23 years. All the subjects participating in this experiment met the mini-
mum criteria of acuity of 20:30 vision, stereo-acuity at 40 sec-arc, and passed a
color vision test. The participants were not aware of the purpose of this experi-
ment or that one of the stereo-pair images was undistorted. Before starting the ex-
periment, all the subjects received instructions and completed a training trial of the
stereo display. This training trial contained four sets of stereo-pair images viewed
and rated in a same way to that in the actual test, but the results from the training
trial were not included in the result analysis. The observers then completed the ex-
perimental trials for each distortion category. They were asked to be as accurate as
possible to judge the image quality. The experiment lasted 50 minutes including
short breaks after each distortion type.
In each trial the images were rated on a sliding scale of Excellent, Good, Fair,
Poor, and Bad. The participants were asked to assess the overall quality of each
stereo-pair images by filling in an answer sheet. After the quality evaluation for all
the images was finished, difference mean opinion scores (DMOS) were calculated
using ITU-R Recommendation BT.500 on a scale of 0-100 as follows. Raw scores
were firstly converted to raw quality difference scores as following:
DR
=
−
R
(1)
ij
iref
()
j
ij
where
i
R
denotes the raw score of the
j
-th stereoscopic image given by the
i
-th
subject, and
iref
R
denotes the raw quality score assigned by the
i
-th subject to
the reference image against the
j
-th distorted image. Then, the raw difference
scores
()
i
D
were converted into
DMOS
j
by averaging the raw difference scores.
2.2 Experimental Results and Analysis
Two phenomena were observed in the experiment. Firstly, the luminance of the
stereoscopic display device can affect the eye strain when subjects evaluated im-
age quality. The dominant eye will feel uncomfortable when the luminance values
