Technology advances aimed at creating a better migration path for 3DTV are rapidly undermining content distributors' earlier assumptions about how to proceed and, in the process, raising the risks of choosing one approach over another.
Until recently, it appeared that a starting point pegged to frame-compatible technology, where video feeds for each eye are delivered at half resolution over a 2D HDTV channel, was the obvious choice (see January issue, p. 1). A second phase employing a full-resolution service-compatible approach tied to MPEG-4 H.264 compression and requiring about one and a half times the bandwidth of a 2D MPEG-4 channel would kick in when service providers were ready to turn up MPEG-4 set-tops on a massive scale.
In this second-phase approach, two separate streams for alternating frames are sent, one containing all the information common to frames for both eyes plus the extra information for one eye and the other transmitting just the extra information for the other eye. However, proprietary iterations of the frame-compatible encoding processes on offer from the leading players, including RealD, Sensio, Dolby and NHK Media Technology, have greatly complicated the decision-making process for phase 1, not only as to which frame-compatible technology to ride but also as to whether operators would need to go to an interim phase 2 service-compatible approach at all.
These vendors say the improvements built into their frame-compatible solutions are sufficient to meet quality requirements until service providers are ready to support the ultimate service-compatible approach, which would be delivery of complete full resolution streams for each eye at higher frame rates. And they say their encoders operating the frame-compatible solution today are designed to support the full service-compatible approach without changes in hardware.
Meanwhile, advances attending the service-compatible approach are raising the possibility that going through the frame-compatible phase will be less bandwidth efficient and more costly in the short run and far less smooth as a transition step to the ultimate phase 3 scenario. Making this case against RealD and the other leading forces in half-resolution frame-compatible technology is TDVision Systems, Inc., which, while much smaller and lacking the market clout of the others, happens to be the patent holder on much of the technology underlying the Multiview Video Coding (MVC) extension to H.264 that anchors the Blu-ray Disc Association's new 3DTV standard.
"We're talking with MSOs to let people know they don't have to settle for one-half resolution," says Ethan Schur, chief marketing officer at TDVision. Schur and his colleagues were on hand to make their case at the CableLabs Summer Conference in August.
"We knew from the beginning that half-resolution and non-compatible formats for cable were only interim solutions," says Manual Gutierrez Novelo, TDVision CEO and CTO and inventor of the company's 2D+Delta technology. "Our system provides higher quality at lower cost."
This claim rests on the fact that 2D+Delta delivers the full-resolution stream created in the 3D production process for the left eye plus just the increment of information that needs to be added to create the right-eye view, eliminating duplicative delivery of the shared information. This cuts total bandwidth by a little over 50 percent compared to what would be required to deliver the complete full-resolution frame for the right eye.
The 3D set-top decoders retrieve the 2D and the "delta" to reconstruct the full HD 3D stereoscopic video while legacy decoders and set-top boxes ignore the 3D information and display the full 2D HD signal. Thus a single encoding process at the headend provides a single feed that can be viewed as 2D or 3D, depending on how the end user is equipped. Or, as Schur puts it, "You encode once, deploy anywhere."
The upshot is that, while the 2D+Delta approach requires about 35 percent more bandwidth for a 3D channel than the frame-compatible approach, which requires an extra 10-15 percent over what's needed for a 2D HD channel, the former would accommodate the 2D feed, eliminating the need for channels dedicated to each. "We consume 45 percent more bandwidth compared to a single 2D channel to deliver 2D and 3D whereas you require 115% more bandwidth and a double production pipeline to deliver 2D and 3D versions of the same program using frame compatible for 3D," Schur says.
But there's a cost tradeoff insofar as the 3D version of the content delivered over the channel will require installation of a new set-top, in contrast to the frame-compatible systems, which can be deployed on existing MPEG-2 2D set-tops that have sufficient processing power to support software upgrades. On the other hand, MPEG-4 is an important advantage which service providers must embrace in any event, Schur notes. "We think 3D will push forward adoption of MPEG-4," he says.
He also argues that only purists will notice or care about the fact that in the 2D+Delta approach the 2D feed is taken from the left-eye component of the stereoscopic camera intake. Left- and right-eye shots are sufficiently close together in 3D production to make the difference between a left-eye 2D view recorded by a 3D camera and a 2D camera view virtually indistinguishable, he says.
TDVision achieved an important milestone in early September when system-on-a-chip maker Sigma Designs became the first SoC supplier to publicly support the TDVision system. "By integrating TDVision's technology into our media processors, we are enabling full HD 3D decoding over IPTV, satellite, cable and over-the-air devices," says David Lynch, vice president and general manager of Sigma's Media Processor Division. The chips will decode streams to fit the user's device and bandwidth environments using different profiles, including MVC, enhancements proprietary to TDVision and hybrid implementations that combine MPEG-2 for the 2D left-eye component and H.264 for the right-eye delta.
This last combination is vital to the legacy 2D pitch TDVision is making for 2D+Delta. "We can tweak the MPEG-2 down a bit and make room for MPEG-4 [within the allocated channel space]," Schur notes. While 3D viewers would require new set-tops, this approach allows the operator to continue serving the legacy set-tops while leveraging the existing channel stream to support full-resolution 3DTV.
To accomplish such encoding tricks, over a year ago TDVision teamed with Magnum Semiconductor, a supplier of video encoding and processing systems, to develop a full HD 1080p stereoscopic 3D encoder supporting MVC, H.264 and MPEG-2 for use in various combinations on any given channel stream. In January Magnum said it was offering a playback and recording solution based on the new encoding system.
Such developments are evidence that "the industry is moving from half resolution frame-compatible interim formats to our service-compatible and full HD 3D solutions," says Gutierrez-Novelo. But, judging from the confusion swirling around all the new developments, this might be a bit of an over statement.
Indeed, says Mark Dzuban, president and CEO of the Society of Cable Telecommunications Engineers, the state of affairs merits caution on the part of the cable industry when it comes to determining a collaborative migration path to 3D, notwithstanding his organization's launch of a "3D-over-Cable" project last year. CableLabs, too, is pursuing a 3DTV initiative, which should help the industry reach consensus on strategies that will determine what changes in transport and other standards must be made at the SCTE.
"Our intent is to find consistency and come up with some answers," Dzuban says. "We're walking through this with the industry with our eyes wide open. But I'm a three- to five-year technology planner, and I'm always careful not to commit prematurely. There's just not enough content or enough unity on technology to get to scale at this point."
But with DirecTV having launched its three-channel 3D service and a growing number of 3D programming channels in the offing as well as ever more events, the cable industry and other providers are under pressure to make choices. Given the challenge posed by TDVision and its allies to the frame-compatible agenda, the question is whether the original track envisioned by cable operators is the best one and, if so, does it make sense to uniformly embrace one of the frame-compatible options as the preferred cable industry path, as DirecTV has done with RealD?
The difficulty here lies in the fact that each of the leading frame-compatible competitors has come up with its own approach to minimizing the quality degradation that results from eliminating pixels to squeeze left- and right-eye views into each frame. Where the original approaches produced a combined 3D resolution barely on par with standard definition, not to mention additional drawbacks when it came to the viewing experience with shutter glasses, the new solutions are designed to at least partially overcome these drawbacks.
As reported in January (p. 8), the RealD system, as demonstrated by DirecTV at the Consumer Electronics Show, eliminates the choppiness in fast-action sequences that was prevalent with earlier iterations of shutter-viewed frame-compatible content. And the picture quality is superior to earlier iterations of side-by-side as well, thanks to a proprietary encoding process that preserves different mixes of pixels for each eye, thereby reducing the net loss.
Sensio takes a somewhat different approach by sampling left and right source images in a checkerboard pattern for purposes of eliminating pixels on the diagonal, where viewer sensitivity to degradation is lower than on horizontal or vertical alignments. In the decompression process Sensio interpolates dropped pixels with degradation-free pixels, resulting in a display output that, to the viewer, comes very close to full resolution.
While most major manufacturers have aligned with RealD, Vizio has embraced Sensio, and some manufacturers are said to be looking at embracing other decoding systems, including Dolby's new open 3D specification and NHK's MT, possibly in dual combinations. All of these solutions include techniques in the encoding process that will allow the same encoders to deliver full-resolution solutions when the market is ready.
Further complicating matters, if 3D becomes a prevalent mode of TV viewing, as some people contend it will, content providers will have to determine whether it makes sense to do productions in both 2D and 3D or to simply go to 3D production for all shows while using one feed to serve as the 2D version for non-3D viewers. This choice would obviously argue for taking the 2D+Delta/MVC approach to distribution, although it should be noted that where live sports programming is concerned, separate 2D and 3D productions will likely be necessary owing to the significant differences in camera placements required for 3D coverage.
While early distributors of 3D programming have lined up behind the frame-compatible technology, there are signs the market could shift toward the service-compatible mode. In October South Korea broadcaster KCC plans to begin a trial broadcast in some regions of the country using MPEG-2 coding for the left-eye stream and H.264 coding for the right stream, all within the allocated 6 MHz channel band.
The idea of being able to deliver both 3D and 2D over the same channel is catching on elsewhere as well. As reported in the online European journal Broadband TV News, a survey conducted by the European Broadcasting Union found that 52 percent of EBU members prefer the service-compatible mode over frame-compatible. Only 11 percent lean toward frame compatible, while about 26 percent believe none of the systems currently on offer meet their needs.