Image Processing Reference
In-Depth Information
techniques, the effects of artifacts introduced by the transmission system have not
received much attention compared to the 2D image/video counterpart. Some of
these artifacts influence the overall image quality, for instance blurriness, lumi-
nance, and contrast levels, similar as in 2D image/video. The effect of transmission
over band-limited and unreliable communication channels (such as wireless chan-
nels) can be much worse for 3D video than for 2D video, due to the presence of two
channels (i.e., stereoscopic 3D video) that can be impaired in a different way; as a
consequence the 3D reconstruction in the human visual system may be affected.
Some networks introduce factors directly related to temporal domain
de-synchronization issues. For instance delay in one view could lead to temporal
de-synchronization and this can lead to reduced comfort in 3D viewing.
The methods employed to mitigate these artifacts (e.g., error concealment) need
to be carefully designed to suit 3D video applications. The simple application of 2D
image/video methods would not work effectively for 3D image/video, as discussed
in [ 4 ] for different error concealment algorithms for 3D video transmission errors.
In [ 4 ], it is observed that in some cases switching back to the 2D video mode is
preferred to applying 2D error concealment methods separately for left and right
views to recover missing image information during transmission. There could be
added implications introduced by these artifacts into our HVS. Therefore artifacts
caused as a result of 3D video streaming can be clearly appreciated only by
understanding how our HVS perceives different 3D video artifacts.
The HVS is capable of aligning and fusing two slightly different views fed into
the left and right eyes and hence of perceiving the depth of the scene. Both
binocular and monocular cues assist our HVS to perceive different depth planes
of image objects [ 5 ]. Binocular disparity is the major cue used by the HVS to
identify the relative depth of objects. Other monocular depth cues include perspec-
tive, occlusions, motion parallax, etc. ([ 5 ]). During 3D video streaming, one view or
both views could be badly affected by channel impairments (e.g., bit errors and
packet losses caused by adverse channel conditions, delay, jitter). For instance,
frame freezing mechanisms employed to tackle missing frames caused by trans-
mission errors or delay could lead to temporal de-synchronization where one eye
sees delayed content compared to the other eye. There are two implications
associated to the case where one view is affected by transmission impairments;
￿ Binocular suppression
￿ Binocular rivalry
Our HVS is still capable to align and fuse stereoscopic content if one view is
affected by artifacts due to compression, transmission, and rendering. Binocular
suppression theory suggests that in these situations the overall perception is usually
driven by the quality of the best view (i.e., left or right view), at least if the quality
of the worst view is above a threshold value. However this capability is limited and
studies show that additional cognitive load is necessary to fuse these views
[ 6 ]. Increased cognitive load leads to visual fatigue and eye strain and prevents
users from watching 3D content for a long time. This directly affects user percep-
tion and QoE. If one of the views is extremely altered by the transmission system,
Search WWH ::




Custom Search