Compression - 3DTV: Processing and Transmission of 3D Video Signals - page 75

note that the disparity-compensated prediction scheme is not applicable to
video-plus-depth type 3D video, since the texture characteristics of colour
and depth videos are different.
Asymmetric [13] and mixed-resolution coding techniques are also available
to left-eye- and right-eye-based stereoscopic video coding, based on the
suppression theory of human visual perception. The suppression theory
of human visual system states that the overall perception of the stereo-
vision is primarily affected by the highest resolution of the video frame
sequences arriving at either eye [14]. Accordingly, a significant amount of
bits can be saved by encoding one of the viewpoints either with reduced
spatio-temporal resolution or with lower fidelity. Such coding approaches
are further applicable to video-plus-depth type 3D video and accordingly,
the depth map transmission overhead can be maintained at around 20% of
that of the colour texture video with pleasant output 3D video quality.
MPEG-4 Multiple Auxiliary Component (MAC) [15] allows image
sequences to be coded with a Video Object Plane (VOP) on a pixel-by-pixel
basis, which contains data related to video objects such as disparity,
depth and additional texture. However, this approach needs to send the
greyscale/alpha shape with any other auxiliary component (e.g. disparity)
and as a result, the coding efficiency is negatively affected.
One of the major challenges in designing compression suits for stereoscopic
video is to carefully analyse the effect of colour perception and depth
perception on the overall stereoscopic video quality, to correctly decide on
the trade-off between colour texture and depth map video coding rates.
Usually, coding depth map information with less accuracy does not affect
the stereoscopic reconstruction quality as much as coding the colour texture
video with less accuracy. Furthermore, due to the characteristics of the
depth image sequence (e.g. large smooth image areas), they can be efficiently
compressed with the existing coding algorithms. Cues of the human visual
system need to be taken into account carefully while assigning different
quantization step sizes, in stereoscopic or multi-view video coding, in order
not to introduce eye-fatigue problems after viewing for a long time.
It is also possible to signal the method(s) used to pack the frames of a
stereoscopic video pair in the video bit-stream, utilizing supplemental user
data (e.g. Supplemental Enhancement Information (SEI) messages defined
in the AVC standard). The AVC standard has already defined such an
SEI message that signals the frame packing method. The Frame Packing
Arrangement SEI message tells the decoder that the left and right view
frames are packed into a single high-resolution video frame (e.g. HD frame
compatible) either in a top-to-bottom, side-by-side, checkerboard, or any
other arrangement. It is also possible to signal the temporal interleaving
scheme used for left and right view frames that are not packed into a single
HD frame (e.g. Full Stereo).
Packing both views into a single video frame makes it possible to use
existing AVC conformant hardware equipment, decoders and set-top boxes