Despite the massive global deployment of high bandwidth networks and the introduction of broadband local loop access technologies, there still exists a place for video compression technology. Bandwidth will always be expensive and the less service providers can use the better. However, whether the current standards for the compression technologies continue to dominate remains to be seen. Neil Brydon, product marketing manager at Harmonic, explains

Compression technology is the foundation stone that drives the business case for digital television. The ability to deliver video successfully over narrow pipes enables businesses such as DTH (direct-to-home) satellite services to be successful. Base-band video originates at 270Mbps and it is both wasteful and impractical to consider moving video at this rate. Yet, it is now commonplace to deliver MPEG-2 (moving picture expert group) video compressed by a factor of over 100:1 at quality levels acceptable to the consumer.

Today, consumers receive useable digital video over satellite, cable or through their copper phone lines. However, these delivery applications have very different compression requirements:

· A cable network has low ongoing costs and offers plenty of bandwidth which means that compression capabilities are not a prime factor in the business model

· Conversely, with satellites bandwidth is precious. They have a finite life and are expensive to deploy, so compression plays a key role for using the bandwidth effectively to send a bouquet of programmes into the home

· The telecom scenario is different again. The twisted pair copper phone lines were only designed for voice. But new modulation techniques coupled with high performance compression do enable the telecom operators to squeeze a video programme into the home.

Each of these applications requires a compatible set-top-box (STB) to decode the incoming video and present it to the TV using a standard that is 50 years old. This highlights a key point: established consumer standards present high inertia and strong resistance to change.

With age comes respect
There are frequent articles and demonstrations of alternative compression technologies including: Wavelet, Fractal, MPEG-4 and H26x. Thus far, however, they have failed to usurp the position in the market held by the dominant standard of MPEG-2.

MPEG-2 has been a phenomenal success and is a great example of global technical cooperation. The standard is mature, and it enjoys huge economies of scale that helps to sustain its leading position.

One reason for this is that it is supported by a vast array of cost-effective silicon, set top boxes and support infrastructure, which has taken time and effort to evolve. Cost is king in the consumer kingdom and it has helped that the MPEG-2 was deliberately designed around a complex encoder and a simple, relatively low cost, decoder. This infrastructure is further reinforced in the professional domain with well-proven MPEG-2 headend infrastructure and effective conditional access technology to ensure successful revenue generation. Furthermore, some of the early criticisms raised against MPEG-2 have been solved.

One complaint was the difficulty associated with managing the compressed signals, but this problem has now been overcome. Stream processors are now available to re-purpose the compressed signals to enable ad insertion, splicing, rate changing and logo insertion - all in the digital domain.

The Challengers
Wavelet has been used for some niche applications, but the process is very symmetrical with complex decoding. Fractals are another interesting compression technique, but this has not progressed beyond the academic domain. Of primary interest are the alternative standards that potentially offer performance advantages over MPEG-2 techniques.

Yet, a key factor is the complexity and cost of decoding alternative standards. Currently, the processing power and memory requirements appear prohibitively expensive but, of course, Moore's Law will ensure that current costs will come down. There is no doubt that MPEG-2 will be superseded, the question is 'when'?

MPEG-4 was developed specifically to extend the abilities at low bit rates and there is some overlap between MPEG-4 and MPEG-2. There are also some big differences. MPEG-4 offers some object-based processing tools that are great for processing synthetic content. The big challenge, however, is whether object-based processing can also produce practical benefits for natural video applications.

Nevertheless, there are other tools within the MPEG-4 video verification model that offer to stretch video capabilities further. Moreover, MPEG-4 continues to increase its encoding capabilities while simultaneously overcome any decoding challenges. At some point the sweet spot may be reached and allow this standard to attain critical mass and threaten MPEG-2.

The H26x standards have developed from a telecom direction through research into video telephony and observers believe that these standards may exceed the capabilities for video of MPEG-4. The standards offer tools that relate strongly to standard MPEG-2 techniques and add complexity to generate additional gain. For example, bi-directional prediction beyond the adjacent frames, motion compensation that uses multiple reference blocks or variable sized blocks.

Growing stronger Many of the supporters of alternative standards forget that that there are still gains to be made in the MPEG-2 domain. MPEG-2 technology has improved fairly linearly. Since the standard was deployed in 1995, the leading operators have been delivering broadcast quality video with around 15% fewer bits per programme each year. There is good reason to believe that this can continue for some time. Pushing MPEG-2 further must occur within the standards framework, so it is unlikely that there will be any radical breakthrough. Advances are made with solid focus and research to refine existing techniques combined with a useful measure of technical innovation.

MPEG-2 offers numerous tools and encoding options. The art of good encoding is to use these tools and MPEG-2 resources as appropriately as possible. The GOP (group of pictures) structure is highly dynamic, I/P/B rate allocation and encoding strategy varies dramatically depending on the type of incoming content: film, video, fast, slow, still, text and so on.

The position of I frames is dependent on scene cut information; difficult scenes are detected and appropriate action taken, for example, when dissolves and fades are recognised. Text detection is applied to help the encoder find and process text more effectively.

Brightness adaptation is used to bias the intra-picture allocation to reduce artefacts that become evident in dark content. Some techniques simply require serious amounts of processing power; for instance leading encoders will apply both DCT (discreet cosine transform) scan modes on a picture-by-picture basis and select the most efficient one. Border processing is another powerful technique to allow fewer bits to be spent on the outer areas of the picture.

Existing techniques offer scope for substantial refinement, plus other innovative techniques have been demonstrated to push MPEG-2 further. There is certainly growing interest to reinforce MPEG-2 at the lowest (sub 2Mbps) rates. Texture discrimination offers to process areas of the picture selectively, and 'smooth' areas of the picture that would otherwise not compress well.

New modes have been demonstrated that use de-interlacing techniques to bring some of the advantages seen with progressive modes. This mode offers to trade fewer compression artefacts against the effects of frame filtering. Advanced versions of lower resolution modes such as SIF (352x240 resolution) are also likely find a useful role, maybe in xDSL video delivery applications.

Noise reduction
All content tends to contain some noise, the random, unwanted and spurious variation in a signal. Even modern facilities equipped with the latest digital production equipment will inevitably import some noisy content. Content such as archive material, film and news feeds, is likely to contain high levels of noise. Noise does not compress well and video compression systems must waste bits to reproduce the noise present in the signal. Noise reduction (NR) and filtering can substantially improve the video received by the viewer - if the right techniques are applied to remove noise prior to compression.

Selectively removing noise is a challenge because the noise shares the same space as valuable picture detail. An ideal noise reduction process will allow powerful suppression of random noise while preserving clean video content.

Good noise reduction means applying filters that preserve edge structure and avoids annoying blurring or trailing effects. Motion Compensated noise reduction is making a significant impact on the DTV industry by saving bits and allowing operators to deliver consistently higher quality video at lower rates than ever before. The major advantage of MCTF (motion compensated temporal filtering) is its inherent ability to remove noise without introducing motion blur artefacts.

Bandwidth boost
Of course, compression will not be required in an end-to-end architecture where bandwidth and storage are virtually free. Fibre all the way to the home is an expensive luxury at the moment. Typically it costs three times more per home than the cost of a hybrid fibre coax installation.

Despite this, some forward thinking cities are helping to subsidise fibre installations. However, even if the last mile is broadband, there are currently other bandwidth constraints in the system, and so compression will be needed until these problems are overcome. Some residences may never get economic broadband connections and these homes will rely on high performance compression application over DBS, xDSL or DVB-Terrestrial for video delivery.

MPEG-2 is very strongly entrenched and vendors are sure to support the current business cases that are built around the existing MPEG-2 deployed infrastructure. There are also problems associated with deploying any new consumer equipment. This will ensure that when change does come it will not happen overnight.

However, alternative standards are likely to be closely related to existing technologies. Smart vendors will have been watching the scenarios unfold, and be ready to launch alternative products when the market is primed for the next technology advancement.

This is a classic chasm scenario; any new technology must be substantially better and meet tough economic criteria to force a switch from their existing solution. MPEG-2 has the high ground today. Other potential alternatives are emerging and being demonstrated. The most likely scenario is that a new standard will establish itself in a niche application before making a concerted challenge on the mainstream MPEG-2 positions.

There are a number of competing compression technologies vying with MPEG-2 for market-share. However, MPEG-2 products, services and solutions are well entrenched in customer locations. Today, these MPEG-2 solutions cost effectively and efficiently satisfy customer requirements. In order for MPEG-2 to be replaced, a new technology must prove it can seriously compete with MPEG-2. Meanwhile, MPEG-2 solutions providers are finding ways to enhance their products and breathe more life into this mature compression technology.