Although 3-D content is widely available in theaters, and 3-D disc players are now on retail shelves, terrestrial digital television has not caught up yet, and a standard supporting 3-D transmission is not yet in the books. Last year, the ATSC concluded a study of 3-D television, with a goal of producing a report on the benefits and limitations of a standard or a set of standards for terrestrial delivery of 3-D television. The report covered various elements of 3-D, including visual sciences and technology. This article will examine various solutions that may become part of a future terrestrial standard for 3-D transmission.
Independent display coding
In one option, the 3-D program can be independent and different from the 2D program. These types of 3-D transmission architectures fall into two large classes: those in which the 3-D program is transmitted alongside a separate, essentially identical, 2D program, and those in which the 2D and 3-D programs are different productions of the same program.
When 3-D and 2D content are coded independently, this can be called MPEG-2 Dual; the 2D view is separately coded, plus independent left and right views are coded, all using MPEG-2 coding. Basically, broadcasters are coding three different versions of the same program, and transmitting them in the multiplex. Alternatively, MPEG-2 MVC (Multi-view Coding) can be used for the 3-D program, using inter-view prediction. This latter option uses coding tools, in which one view is coded in a main-profile (MP) base layer, and that layer is used to predict the other view in an enhancement layer with temporal scalability tools. However, MPEG-2 MVP is not expected to offer a significant coding gain over the independent encoding of the two views, owing to the limitations of MPEG-2 coding efficiency.
Frame-compatible 3-D is an architecture in which the left and right views are decimated (usually by a factor of 2) and arranged into a frame-compatible format such as side-by-side or top-and-bottom. These frame-compatible formats fit into a conventional 30Hz frame period, so no additional baseband bandwidth is needed, and conventional baseband video equipment can route the signals. (Of course, the spatial resolution at the home receiver is compromised with respect to a 2D program.) Frame-compatible video can be encoded using a conventional MPEG-2 or H.264 Advanced Video Coding (AVC) codec, but it cannot be displayed on a 2D display, which cannot separate or properly integrate the left and right views. Again, the independent 2D program is separately coded using MPEG-2.
A variation of frame-compatible 3-D is full-resolution frame-compatible 3-D, in which the same frame-packing arrangement is used to encode the left and right views, but no sub-sampling is performed. Thus, a higher baseband bandwidth is required at the codec I/O, with an associated decrease in efficiency compared with inter-view predictive coding. In this format, the 3-D program is most likely coded using AVC, as MPEG-2 could not handle the increased bandwidth, especially with the 2D program also present in the multiplex.
MVC 3-D is a format in which the left and right views are coded at full resolution, either independently using AVC or jointly using MVC with inter-view prediction. The latter approach can result in a lower bandwidth requirement than independent coding. But it comes with a hardware tradeoff between an architecture supporting two compression streams, and two parallel codecs, versus one more-complex stream.
An alternative to MVC with inter-view prediction is AVC frame compatible with resolution enhancement, in which the base layer carries the left and right views in a frame-compatible format, and the enhancement layer carries the “difference information” needed to provide full resolution. This format provides a migration path that would enable 3-D receivers to be built today, using existing AVC tools for the base layer. Later receivers would add full resolution, using AVC tools not yet developed. Those early receivers would thus be compatible with, but not capable of displaying, the future format.
Dependent display coding
The other option is for the 3-D program to be dependent and related to the 2D program. When one view is used as the 2D program, and both views provide the 3-D program, shared coding is an option in which either both views are coded independently, or MVP is used to predict one view from the other. (See Figure 1.)
There is some sensitivity in the production community that this type of coding arrangement constrains the director's ability to shoot the scene because one view is subservient to the other. Nonetheless, it can provide savings in both production budgets and transmission bandwidth.
Again, as with the independent formats, both views can be coded using MPEG-2, or a hybrid system could be utilized that encodes the “default” view using MPEG-2 and the second view with AVC or another advanced codec, such as high-efficiency video coding (HEVC). With this architecture, one consideration deserving careful scrutiny is the fact that the left and right views are coded by different means, a point that applies to other formats as well. The result can be different compression artifacts in the two views, causing some peculiar effects. Demonstrations of hybrid architectures suggest this could be acceptable with further study.
An alternative to coding left and right views is the use of a depth map, a grayscale video picture in which the depth of an object in the image is coded by using different intensity levels. The depth map then is coded as an ordinary video stream, using AVC or another advanced codec. Transmission variations include combinations of the techniques discussed previously.
Depth information must be extracted (or synthesized) from original left and right views, or from temporal information derived from moving objects or a moving camera. This 2D-plus-depth-map technology is still in an early stage, so error-free depth information is not yet achievable. In addition, the decoder must re-generate the second view, which adds computational complexity. Nonetheless, the approach has interesting applications, especially for animation and graphics.
As the solutions studied in the ATSC report must be backwards-compatible with existing MPEG-2 receivers, all the variations considered include a standard 2D-content channel using MPEG-2 coding. But such a constraint likely will be lifted somewhere down the road, when MPEG-2 eventually is made obsolete by AVC and HEVC.
In the meantime, a system can be developed that fits into existing transmission standards, and that means that consumer products could be available in 2013. A full analysis of the pros and cons of each of the discussed approaches can be found in the Planning Team 1 Report on 3-DTV at the ATSC website, www.atsc.org.
Aldo Cugnini is a consultant in the digital television industry.
Send questions and comments to: firstname.lastname@example.org