Information Technology Reference
In-Depth Information
Although these objective quality metrics have been validated to be accurate in
evaluation of quality degradations for normal size images [21], there is no guaran-
tee that they will perform as well for a digital cinema system because of the dif-
ference between conventional visual presentations and visual presentation in a
digital cinema setup. To our best knowledge, no rigorous analysis of the perform-
ance of objective image quality assessment specific to digital cinema has been
studied so far. Here, we will study the difference of image quality assessment be-
tween a normal size image and a high-resolution image displayed on a large
screen, from the viewpoint of the HVS. We will adopt related characteristics of
the HVS to analyze this difference and propose an HVS-based approach to im-
prove the performance of PSNR and SSIM in image quality assessment for digital
cinema. In addition, the human vision cannot perceive all distortions in an image,
especially in those regions with high-frequency, because of the contrast sensitivity
of the HVS that varies with different frequencies, and the existence of masking ef-
fects [22]. Therefore, a modified PSNR by taking into account the contrast mask-
ing effect and removing the imperceptible distortion from the quality computation
has been proposed in [23] and will be employed in this study. The experimental
results with respect to the subjective quality evaluation demonstrate that the pro-
posed approach can evidently improve the performance of existing metrics in
image quality assessment for digital cinema systems.
4.1 Visual Characteristics in D-Cinema and Its Application to
Image Quality Metrics
The most significant difference between digital cinema images and traditional im-
ages is that the former have a much higher resolution and are shown on much larger
screens. Therefore, the quality assessment of the image presentation in a digital
cinema setup is accordingly different from that in other controlled laboratories on a
normal size display, such as a TV or computer monitor.
In the human visual system, eye movement is typically divided into fixation
and saccades. Fixation is the maintaining of the visual gaze on a single location.
Saccade refers to a rapid eye movement. Humans do not look at a scene in fixed
steadiness, instead, the fovea sees only the central 2º of visual angle in the visual
field and fixed on this target, then moves to another target by saccadic eye move-
ment [24]. Saccades to an unexpected stimulus normally take about 200 millisec-
onds to initiate, and then last about 20-200 milliseconds, depending on their
amplitude (20-30 milliseconds is typical in reading). In image quality assessment,
quality evaluation takes place during eye fixation when the fovea can perceive the
visual stimulus with maximum acuity [25]. Thus, when viewing an image on a
large screen in the digital cinema, subjects cannot see the entire image at once and
evaluate distortions in all regions. Even though the fovea might not be able to per-
ceive an entire image simultaneously on a normal size display, we believe the
situation in this case is significantly different from that in a digital cinema setup.
Many physiological and psychological experiments have demonstrated that
human attention is not allocated equally to all regions in the visual field, but
