We start with a generic discussion about transmission approaches and then look at DVB-based satellite approaches.
Overview of Basic Transport Approaches
It is to be expected that 3DTV for home use will likely first see penetration via stored media delivery (e.g., Blu-ray Disc). The broadcast commercial delivery of 3DTV (whether over satellite/DTH, over the air, over cable, or via IPTV), may take a few years because of the relatively large-scale infrastructure that has to be put in place by the service providers and the limited availability of 3D-ready TV sets in the home (implying a small subscriber, and so small revenue, base). Delivery of downloadable 3DTV files over the Internet may occur at any point in the immediate future, but the provision of a broadcast-quality service over the Internet is not likely in the foreseeable future.
There are a number of alternative transport architectures for 3DTV signals, also depending on the underlying media. The service can be supported by traditional broadcast structures including the DVB architecture, wireless 3G/4G transmission such as DVB-H approaches, Internet Protocol (IP) in support of an IPTV-based service (in which case it also makes sense to consider IPv6), and the IP architecture for Internet-based delivery (both non–real time and streaming).
The specific approach used by each of these transport methods will also depend on the video-capture approach, as depicted in Table 4.1. Initially conventional stereo video (with temporal multiplexing or spatial compression) will be used by all commercial 3DTV service providers; later in the decade other methods may be used. Also, make note in this context that in the United States one has a well-developed cable infrastructure in all Tier 1 and Tier 2 metropolitan and
suburban areas; in Europe/Asia, this is less so, with more DTH delivery (in the United States DTH tends to serve more exurban and rural areas). A 3DTV rollout must take these differences into account and/or accommodate both.
Note that the V + D data representation can be utilized to build 3DTV transport evolutionarily on the existing DVB infrastructure. The in-home 3D images are reconstructed at the receiver side by using DIBR. MPEG has established a standardization activity that focuses on 3DTV using V + D representation.
There are generally two potential approaches for transport of 3DTV signals: (i) connection-oriented (time/frequency division multiplexing) over existing DVB infrastructure over traditional channels (e.g., satellite, cable, over-the-air broadcast, DVB-H/cellular), and (ii) connectionless/packet using the IP (e.g., “private/dedicated” IPTV network, Internet streaming, Internet on-demand servers/P2P i.e., peer-to-peer). These references, for example, among others, describe various methods for traditional video over packet/ATM (Asynchronous Transfer Mode)/IPTV/satellite/Internet [1–5]; many of these approaches and techniques can be extended/adapted for use in 3DTV.
Figures 4.1–4.7 depict graphically system-level views of the possible delivery mechanisms. We use the term “complexity” in these figures to remind the reader that it will not be trivial to deploy these networks on a broad national basis.
A challenge in the deployment of multi-view video services, including 3D and free-viewpoint TV, is the relatively large bandwidth requirement associated with transport of multiple video streams. Two-streams signals CSV, V + D, and LDV are doable: the delivery of a single stream of 3D video in the range of 20 Mbps is not outside the technical realm of most providers these days, but to deliver a large number of channels in an unswitched mode (requiring say 2 Gbps access to a domicile) will require FTTH capabilities. It is not possible to deliver that content over an existing copper plant of the xDSL (Digital Subscriber Line) nature unless a provider could deploy ADSL2+ (Asymmetric Digital Subscriber Line; but why bother upgrading a plant to a new copper technology such as this one when the provider could actually deploy fiber? However, ADSL2+ may be used in Multiple Dwelling Units as a riser for a FTTH plant). A way to deal with this is to provide user-selected multicast capabilities where a user can select an appropriate content channel using IGMP (Internet Group Management Protocol). Even then, a household may have multiple TVs (say three or four) switched on
simultaneously (and maybe even an active Digital Video Recorder or DVR), thus requiring bandwidth in the 15–60 Mbps. MV + D, where one wants to carry three or even more intrinsic (raw) views, becomes much more challenging and problematic for practical commercial applications.
Off-the-air broadcast could be accomplished with some compromise by using the entire HDTV bandwidth for a single 3DTV channel—here, multiple TVs in a household could be tuned to different programs.
However, a traditional cable TV plant would find it a challenge to deliver a (large) pack of 3DTV channels, but it could deliver a subset of their total selection in 3DTV (say 10 or 20 channels) by scarifying bandwidth on the cable that could
otherwise carry distinct channels. The same is true for DTH applications.
For IP, a service provider–engineered network could be used. Here, the provider can control the latency, jitter, effective source–sink bandwidth, packet
loss, and other service parameters. However, if the approach is to use the Internet, performance issues will be a major consideration, at least for real-time services. A number of multi-view encoding and streaming strategies using RTP (Real-Time Transport Protocol)/UDP (User Datagram Protocol)/IP or RTP/DCCP (Datagram Congestion Control Protocol)/IP exist for this approach. Video streaming architectures can be classified as (i) server to single client unicast, (ii) server multicasting
to several clients, (iii) P2P unicast distribution, where each peer forwards packets to another peer, and (iv) P2P multicasting, where each peer forwards packets to several other peers. Multicasting protocols can be supported at the network-layer or application layer [6]. Figure 4.8 provides a view of the framework and system for 3DTV streaming transport over IP.
Yet, there is a lot of current academic research and interest in connectionless delivery of 3DTV content over shared packet networks. 3D video content needs to be protected when transmitted over unreliable communication channels. The effects of transmission errors on the perceived quality of 3D video could not be less than those for the equivalent 2D video applications, because the errors will influence several perceptual attributes (e.g., naturalness, presence, depth perception, eye-strain, and viewing experience), associated with 3D viewing [7].
It has long been known that IP-based transport can accommodate a wide range of applications. Transport and delivery of video in various forms goes back to the early days of the Internet. However, (i) the delivery of quality (jitter-, loss-free) content, particularly HD or even 3D; (ii) the delivery of content in a secure, money-making subscription-based manner; and (iii) the delivery of streaming real-time services for thousands of channels (worldwide) and millions of simultaneous customers remain a long shot at this juncture. Obviously, at the academic level, transmission of video over the Internet (whether 2D or 3D) is currently an active research and development area where significant results have already been achieved. Some video-on-demand services that make use of the Internet, both for news and entertainment applications, have emerged but desiderata (i), (ii), and (iii) have not been met. Naturally, it is critical to distinguish between the use of the IP (IPv4 or IPv6) protocol and the use of the Internet (that is based on IP), as a delivery mechanism (a delivery channel). IPTV services delivered over a restricted IP infrastructure appear to be more tenable in the short term, both in terms of Quality of Service (QoS) and Quality of Experience (QoE). Advocates now advance the concept of 3D IPTV. The transport of 3DTV signals over IP packet networks appears to be a natural extension of video over IP applications; but the IPTV model (rather than the Internet) model seems more appropriate at this time. The consuming public will not be willing, we argue based on experience, to purchase new (fairly expensive) TV displays for 3D, if the quality of the service is not there. The technology has to “disappear into the background” and not be smack in the foreground, if the QoE has to be reasonable.
To make a comparison with Voice over Internet Protocol (VoIP), it should be noted that while voice over the Internet is certainly doable end-to-end, specialized commercial VoIP providers tend to use the Internet mostly for access (except
for international calling to secondary geographic locations). Most top-line (traditional) carriers use the IPover their own internally designed, internally engineered, and internally provisioned network, and also for core transport [8–21].
Some of the research issues associated with IP delivery in general, and IP/Internet streaming in particular, include but are not limited to, the following [6]:
- Determination of the best video encoding configuration for each streaming strategy: multi-view video encoding methods provide some compression efficiency gain at the expense of creating dependencies between views that
hinder random access to views. - Determination of the best rate adaptation method: adaptation refers to adaptation of the rate of each view as well as inter-view rate allocation depending on available network rate and video content, and adaptation of the number and quality of views transmitted depending on available network rate and user display technology and desired viewpoint.
- Packet-loss resilient video encoding and streaming strategies as well as better error concealment methods at the receiver: some ongoing industry research includes the following [7].
- Some research related to Robust Source Coding is of interest. In a connectionless network, packets can get lost; the network is lossy. A number of standard source coding approaches are available to provide robust source coding for 2D video to deal with this issue, and many of these can be used for 3D V + D applications (features such as slice coding, redundant pictures, Flexible Macroblock Ordering or FMO, Intrarefresh, and
Multiple Description Coding or MDC are useful in this context). Lossaware rate-distortion optimization is often used for 2D video to optimize the application of robust source coding techniques. However, the models used have not been validated for use with 3D video in general and FVV in particular. - Some research related to Cross-Layer Error Robustness is also of interest for transport of V + D signals over a connectionless network. In recent years, attention has focused on cross-layer optimization of 2D video quality. This has resulted in algorithms that have optimized channel coding and prioritized the video data. Similar work is needed to uncover/assess appropriate methods to transport 3D video across networks.
- Other research work pertains to Error Concealment that might be needed when transporting V + D signals over a connectionless network. Most 2D error concealment algorithms can be used for 3D video. However,
there is additional information that can be used in 3D video to enhance the concealed quality: for example, information such as motion vectors can be shared between the color and depth video; if color information
is lost, then depth motion vectors can be used to carry out concealment. There are other opportunities with MVC, where adjacent views can be used to conceal a view that is lost. - Best peer-to-peer multicasting design methods are required, including topology discovery, topology maintenance, forwarding techniques, exploitation of path diversity, methods for enticing peers to send data and to stay connected, and use of dedicated nodes as relays.
Some have argued that stereo streaming allows for flexibility in congestion control methods, such as video rate adaptation to the available network rate, methods for packet loss handling, and postprocessing for error concealment but it is unclear how a commercial service with paying customers (possibly paying a premium for the 3DTV service) would indeed be able to accept any degradation in quality.
Developing laboratory test bed servers that unicast content to multiple clients with stereoscopic displays should be easily achievable, as should be the case for other comparable arrangements. Translating those test beds to scalable, reliable (99.999% availability), cost-effective commercial-service-supporting infrastructures is altogether another matter.
In summary, real-time delivery of 3DTV content can make use of satellite, cable, broadcast, IP, IPTV, Internet, and wireless technologies. Any unique requirements of 3DTV need to be taken into account. The requirements are very
similar to those needed for delivery of entertainment-quality video (e.g., with reference to latency, jitter, and packet loss), but with the observation that a number (in not most) of the encoding techniques require more bandwidth. The incremental bandwidth is as follows: (i) from 20% to 100% more for stereoscopic viewing compared with 2D viewing;1 (ii) from 50% to 200% for multi-view systems compared with 2D viewing; and (iii) a lot more bandwidth for holoscopic/holographic designs (presently not even being considered for near-term commercial 3DTV service). We mentioned explicit coding earlier: that would indeed provide more efficiency, but as noted, most video systems in use today (or anticipated to be available in the near future) use explicit coding. Synthetic video generation based on CGI techniques needs less bandwidth than actual video. There can also be content with Mixed Reality (MR)/Augmented Reality (AR) that mix graphics with real images, such as those that use depth information together with image data for
3D scene generation. These systems may also require less bandwidth than actual full video. Stereoscopic video (CSV) may be used as a reference point. Holoscopic/ holographic systems require the most. It should also be noted that while
graphic techniques and/or implicit coding may require a very large transmission bandwidth, the tolerance to nformation (packet loss) is typically very low.
We conclude this section by noting, again, that while connectionless packet networks offer many research opportunities as related to supporting 3DTV, we believe that a commercial 3DTV service will more likely occur in a connectionoriented
(e.g., DTH, cable TV) environment and/or a controlled-environment IPTV setting.