Posted on February 22, 2011 at 19:51 by Kevin Chung in Technology
3D is hot. A marketing machine running like madness. It promises the consumer one of the greatest visual experiences ever! While I’m aware 3D has some nice features, I think the discussion about 3D stands in the way of other -often overlooked- possibilities of the current broadcast and video implementations. I’d like to highlight some of those potential developments here.
I believe that in the current era of Full HD TV broadcasting there’s still plenty of room for improvement. Improvements that should make video higher quality when it comes to human perception. However we still have incorporated things from the past. Take for example MPEG2 and interlacing.
Drop MPEG2 and go H.264/AVC
MPEG2 is still widely used as the end user format for DVDs and also still used by many TV broadcasters. Throughout its existence, some MPEG2 encoders have been vastly optimized to achieve higher quality by improving algorithm efficiency. However H.264/AVC really has the future here. It offers the same quality at a much lower bit rate. Its feature set also allows for more fine tuning of encoding parameters.
Generally I haven’t seen TV broadcasters go beyond 18 Mbit/s if MPEG2 is used for compressing the Full HD video stream. While a highly optimized MPEG2 encoder still delivers good quality pictures you’d still notice blocking on fast motion scenes like live performances. It’s also a bit short in retaining grain. I’d propose 30 Mbit/s to retain fine detail and avoid blocking when using MPEG2. For purists, you will probably want to go all the way up to 70Mbit/s if you care for each grain particle. Of course bandwidth must allow it. Hopefully network infrastructure through cable and satellite will keep on adding faster networks with higher bandwidth.
Until then TV broadcasters could also choose to drop MPEG2 and go for H.264/AVC. I’d encourage them to extensively test their new platform. I’ve seen TV broadcasters making a switch and picture quality was worse in their new setup. That’s of course not supposed to be happen. While there are poor H.264/AVC encoders around, one should pick the good ones and try to optimize it to the requirements of the new platform. Once done, it greatly reduces bit rate needed to achieve a targeted picture quality compared to MPEG2.
Another advantage for operators is bandwidth. Consumers who have a connection that’s far away from the distribution network have a slower throughput of data. But using H.264/AVC still allows them to give these end-users sufficient visual quality where MPEG2 is running short.
Although H.264/AVC is computationally more expensive compared to MPEG2 there are ASIC chips on the market that are highly optimized for decoding 1080p H.264/AVC streams up to 60 fps. From a financial point of view they really aren’t the biggest thing to worry about in new set top boxes. The same applies for hardware encoders.
As any rational human being we should work efficiently. H.264/AVC is less demanding on bit rate and its technical abilities allow details to be better preserved.
Drop interlacing and go progressive
As of many years flat panel TVs show pictures using progressive scan. It’s not logical to still broadcast in 1080i30 (for NTSC areas). These cameras can shoot nicely in progressive scan mode, yet we have to go through the inconvenient process of interlacing and then deinterlacing.
A good example are news anchors. When they wear a suit or shirt with vertical stripes -and it is mostly the male anchors that do so- prepare yourself for some funny outraged lines when deinterlacing is needed for progressive display. Even some sophisticated deinterlacers will have a hard time reconstructing to what is correct to our eyes. This is clearly rather a limitation of field based video. There’s more but I’ll leave it here.
“Interlacing is easy. Deinterlacing is a pain.”
Spatially this hurts details. In practice this doesn’t always need to be bad if sufficient bit rate and a proper deinterlacer is used. But from a theoretical point of view there’s really no point in using interlacing today. Most professionals in the video world would agree on this. Yet we are not controlling this market. Let’s hope they change their mind.
Interlaced (left) vs Progressive (right)
image credits: http://www.hdtvinfoport.com/HDTV.html
The only real advantage would be that 1080i30 is essentially 60 half frames per second. Motion is more smooth. I’d still say interlacing truly has had its time and has become obsolete. The common reason they still use 1080i30 is because it takes much less bandwidth compared to 1080p60. That is true because we are dealing with video that has lost half of its spatial resolution in the case of interlacing.
1080p30 is still an option. Although it is not as smooth as 1080i30 or ideally 1080p60, at least everything stays progressive. Note that modern flat panel TVs nowadays can do a lot more than only showing a picture from the input signal it receives. For example, it can do motion interpolation using different techniques. Not only the engine in a TV itself but also graphics cards from nVidia and AMD can do something similar. For graphics cards this only applies if you are using DXVA hardware decoding. This motion interpolation can give just that extra bit of smooth motion. I’ve seen some pretty good results for film content. Note that film at 24 fps and screen refresh rates of 60 > 120 > 240 > 480 Hz are very incompatible. This is why I had to point it out aside from the fact TVs also have a native 24 Hz mode.
To summarize, stick to progressive:
- It is much easier to apply any post processing filters on. Modern codecs encode much better on progressive video. For example, codecs based on H.264/AVC standard.
- It’s much harder to make fundamentally wrong edits on progressive video than interlaced or telecine video. In fact, if interlacing was not invented there’d probably be no telecine like we still use today.
If these two important aspects have been improved then it’s time to look for something new.
What I desire as the next step for the future?
As I see it, the next step beyond Full HD would be a higher resolution like Quad HD. Compared to SD, Full HD already allowed objects to be better scaled across a frame. For example, a camera panning of a live concert hall where the crowd is better visible in Full HD, rather than making the equation one pixel could only represent one head in the case of SD. With Full HD, objects are relatively better in place and size. Subtitles look much sharper and are larger which improves reading convenience.
However where I do feel Full HD is still lacking, is simply the amount of pixels to describe “fine” detail. Very fine structure from clothing or hair for example needs more pixels. In other words I want higher resolution, more DPI.
NHK’s Super Hi-Vision
As I am writing this, I suddenly remember my experience at IBC 2010. Japanese TV broadcaster NHK presented their Super Hi-Vision technology there: that is video with a resolution of 7680 x 4320 pixels, 16 times (!!!) higher than standard HDTV. There was some very pretty details I could see from their demonstration. It’s no joke, no overstatement. The perceived result was valid and shows they have worked hard to apply this technology to the max. Such words from a videophile. That’s something to dream for!
I don’t know what would be the native resolution the human eye has. The desire is to have that same resolution definition for videos. Just like the fact the human eye doesn’t see motion much smoother beyond 60 frames per second.
I’ll be looking forward to what future technology has to provide and change the way we define video!
Kevin Chung is encoding engineer at ODmedia.
