The Advanced Television Systems Committee (ATSC), the group behind the television standards known by the same initials, today announced that they’ve begun work on ATSC standards for broadcast 3D TV. As announced:
The groundbreaking work on the new 3D-TV broadcast standard builds on the extensive efforts over the last year by the ATSC 3DTV Planning Team. This new standard, which could be completed in a year, will allow:
3D content delivered on one ATSC terrestrial channel to fixed receivers, with delivery of both views (left and right eye) in real-time, and;
3D content delivered on one ATSC terrestrial channel to Mobile/Handheld receivers, and delivery of both views in real-time.
3D content delivered in non-real-time.
It isn’t clear yet just what the final 3DTV standard will entail. Today’s broadcast ATSC for fixed receivers uses MPEG-2 with a maximum resolution of 1080p30. However, back in 2008, MPEG-4/H.264 was added to ATSC with support for up to 1080p60, but this isn’t in use today as main ATSC receivers don’t support MPEG-4/H.264. Most stations are using 1080i or 720p MPEG-2 today.
As with the existing ATSC standards, which give broadcasters numerous choices in resolution, frame rates, etc., the 3D standards will likely do the same. I expect the final standard will give broadcasters a choice of a few different ways to encode 3D video.
As I discussed in my recent post on the state of 3D, 3D broadcasts in the US to date have used a de facto standard known as frame packing. In this system two video frames, left and right eye, are ‘packed’ into one video frame for encoding. So you might take a 1080p frame, which is 1920×1080, and pack two 960×1080 frames into it in left-right packing. Or a 720p frame, which is 1280×720, might have two 1280×360 frames inside in top-bottom packing. The end device handles splitting the frames and displaying them. The obvious drawback is halving the resolution in one direction.
The final standard may include frame packing as one option, albeit a low quality one. I note that backwards-compatibility is mentioned:
“The addition of 3D-TV capability to the DTV broadcast standard will foster new broadcast services while preserving the integrity of legacy TV receivers by adopting a system that allows for simultaneous delivery of 2D HDTV, Mobile DTV, and 3D programs within the same channel while ensuring backwards compatibility,” [ATSC President Mark] Richer said.
To be backwards compatible with today’s receivers, this probably means MPEG-2 using one of three systems – left/right frames, 2D+Delta, or 2D+Depth. Left/right frames is basically frame packing, transmitting two full video frames (one per eye). For 2D display just one frame is used. The drawback is the bandwidth, broadcast ATSC maxes out at 19.4Mbps; by way of comparison Blu-ray supports up to 48Mbps. Sending two full frames means you have to sacrifice the resolution and/or turn up the compression.
So the most likely systems are 2D+Delta or 2D+Depth. Both transmit one full 2D frame, which is displayed on 2D sets. They differ in how they provide the 3D effect. +Delta transmits encoding for how the other frame differs from the full frame that is sent. So if the full frame is the left eye, the encoding is the delta, or difference, to produce the full right eye image – left+delta = right. +Depth instead transmits a ‘depth map’ that overlays the image. This basically tells the display the ‘z-index’ of items in the image. From that it computes the 3D version of the image and generates the left/right images. This is a lot more work for the display device, but can save on bandwidth and allow for higher resolution/quality images. Since Europe standardized 3D for DVB-T on 2D+Depth, it seems likely the US will do the same – or at least include it as one option.
While this would allow broadcasts compatible with all existing MPEG-2 ATSC receivers, it would not provide an optimal 3D experience due to the compromises involved.
Since H.264 isn’t in wide use today, I think there is still a chance for a clean H.264 3D solution. Since 1080p60 2D images are supported, it would be possible to do 1080p30 3D with two full frames – every other frame is left/right. Even better, 1080p30 3D using 2D+Delta/Depth would allow reduced compression in the base 2D image, for a higher quality image.
Since the pipe is the same size for MPEG-2 or MPEG-4, using MPEG-4 will always provide for a better quality image. No matter which video codec or 3D system is used, the 3D image quality will necessarily be lower than plain 2D as you need to encode more data with the same fixed maximum. All else being equal, of course. In other words, if you use the minimum compression required to fit the stream in the pipe the plain 2D image will be less compressed than the 3D image of the same resolution and frame rate as the 3D image needs to fit both the image and 3D data into the pipe.
H.264 will provide the best possible quality; but no matter what they do they won’t be able to match Blu-ray’s 3D quality as it can push full 1080p60 frames to both eyes, with each eye having more available bandwidth than an entire ATSC channel.