Graphics Reference
In-Depth Information
Figure 6.2: Reducedreferenceedgebased3Dvideoqualitymetric [12].
Table 6.2 reports a few existing NR and RR quality metrics for 3D image/video. This table
explainswhichimagefeaturesareusedtomeasuretheoverallperceptionandhowmuchthe
different metrics are correlated with subjective quality scores (i.e., MOS) and with existing
Full-Referencemethods.Itcanbe observed that most of these methods show a high degree
of correlation with subjective MOS and Full- reference methods. However, these metrics
are focused on one or two specific 3D perceptual attributes. The combined effect of these
perceptual attributes which is directly related to user 3D QoE has not been addressed to
date.Themethodsin [128] and [127] areevaluatedusingthesameimagedatabasewhereas
othersareevaluatedusingdifferentdatasets.Sincesomeofthesemetrics,e.g.,NRmetrics
([127] and [129]) are designed for a particular types of image artifacts (e.g., JPEG com-
pression), it is not always possible to compare the performance of a NR metric with anoth-
er objective quality model in a common dataset. On the other hand, due to the overhead
associated with RR metrics compared to zero overhead for NR metrics, the usage and
advantages of these methods are significantly different. In addition, due to some practical
reasons(intellectualpropertyrights,differentsource3Dvideoformats,e.g.,colour+depth
Search WWH ::




Custom Search