Image Processing Reference
In-Depth Information
3.1
Introduction
Telecommunication networks are experiencing large demands for video data,
where IP video traffic is expected to be around 79 % of all consumer Internet traffic
by 2018 [
1
]. These requirements will be even more stressed with the event of
Immersive multimedia services which need the transmission of huge amounts of
data, as multiple camera video content must be transmitted to better render the
three-dimensional (3D) scene at the display. The uptake of this technology will
further push the amount of IP video traffic passing over the networks. This 3D
content has to pass over bandwidth-limited channels and to be stored on space-
limited hardware. Some 3D video content is already being transmitted, with current
technology being limited to stereoscopic video over satellite channels, stored
content on Blu-Ray
TM
disks, and Internet technologies [
2
]. Therefore, in order to
sustain the massive growth in communications needs imposed by the transmission
of 3D video content, adequate video coding techniques are needed. The choice of
the video coding scheme that is used for a particular application or service depends
on the data format and the hardware that is available at the display. This means that
adequate design of video representation methods is needed to facilitate the speed
and amount of compression that can be achieved during 3D video coding.
In this chapter we will first describe the different 3D video representation
formats starting from stereo and then moving to multiview and holoscopic formats.
Emerging formats of multiview video-plus-depth and layered depth video are also
discussed. The chapter then focuses on the different techniques used to encode 3D
videos. These include stereoscopic, multiview, video-plus-depth, and non-scalable
3D holoscopic video (3DHV) coding. A final conclusion is then drawn together
with some discussions on future trends.
3.2
3D Video Representation
This section presents an overview of the existing formats to represent 3D video. All
of these formats aim to represent the captured scene from two or more viewpoints.
Original systems used a stereoscopic video representation, consisting of two views
of the captured scene. When more than two views are used, we refer to the
representation as multiview video (MVV). This allows the representations of the
captured scene from any viewpoint, or its display in autostereoscopic displays.
When the number of used views is very large (hundreds or thousands), we refer to it
as holoscopic video.
Alternative representations may use information beyond the captured views. An
option that has recently gained a great relevance is the use of depth information
associated with a texture view, which is commonly referred to as video-plus-depth.
Recently proposed formats aim to represent more complex 3D models of the
objects in the scene, usually by using a large number of views combined with image
