Harmonic, how it works….and next gen standards…..
The MPEG-2 standard has been hugely successful, delivering broadcast quality video to tens of millions of homes. MPEG-2 will continue to be relevant in many applications for many years to come but it is clear that there is an opportunity developing for a next generation standard to emerge and to be deployed. The next generation standard will be appropriate in markets where greater efficiency is required.
A number of alternative compression technologies have been getting significant attention, the most prominent are MPEG-4, H263, H263+, H263++,H26L. It now appears to be clear that the industry consensus lies with H.26L and that this development standard will become a new highly efficient profile within the MPEG-4 suite. Somewhat confusingly, the H.26L standard when it is complete will be known as H.264 but is also known been known as MPEG-4 part 10 and JVT. H.264 relies on block based encoding techniques with advanced motion estimation and compensation, so the underlying technology is very familiar.
It is foreseen that H.264 will be the first profile that really provides MPEG-4 with a compelling performance advantage for delivery of broadcast quality video. For this application, the MPEG-4 simple profile offered no gain over MPEG-2 and although the advanced simple profile does offer some gains over MPEG-2, the gains have not been sufficiently compelling to get attention in the broadcast arena.
Most of the effort and publicity has centered on proprietary standards that are loosely based around MPEG-4 ASP technology. Examples are Windows Media, RealNetworks, Sorensen,ON2 etc. These have attained some market share for delivery to the PC but have yet to make any impact on delivering interlaced content to the TV.
It has become clear the market required new technology to demonstrate a sufficiently compelling advantage over the incumbent standard. Any new standard also needs to be viable, on an end to end basis on both a technical level and from a business perspective. The prime requirement is for set top boxes that are capable and available at little or no premium to the existing boxes.
Squeezing more out of MPEG
Further improvements to MPEG-2 come as a result of some of the carry back of AVC type techniques and the further use of the additional processing power necessitated by Next Generation Encoding (AVC and WM9)
Harmonic first launched its advanced encoding technique that uses three MPEG-2 processors instead of one when it introduced its LookAhead technology. The LookAhead architecture enjoys advanced visibility of incoming video content, because one encoding processor extracts statistics from incoming video almost one second ahead of the second encoding processor. Such video analysis helps the encoder to:
· Make better encoding decisions
· Make better use of MPEG-2 buffer resources
· Apply improved filter control
The third encoding processor is coupled within the LookAhead control loop to offer motion compensated temporal filtering capabilities.
Lookahead is used in a number of ways
· Powerful control input - about half a second of visibility
· Encoding strategy based on content - see what type of material is coming through, and apply the most appropriate technique that will use the least bits.
· Statistical multiplexing - use Lookahead to understand the complexity of the incoming video, feed this forward to the mux so that an optimal and well time rate allocation is achieved.
More MPEG buffering - One law of MPEG is that the buffer must never overflow or underflow, Lookahead allows the system to safely ride close to the limits (allowing the system to effectively have a bigger buffer and be better able to absorb difficult video).
Dynamic GOP control - Lookahead can identify difficult scenes such as scene cuts, fades, and dissolves (apply special mechanisms to deal with them). A big part of stable MPEG encoding is to correctly balance the rate allocation across the different MPEG picture types …..
Enhanced I/P/B rate allocation - Lookahead stats are used to tell the system where to apply the noise reduction most effectively. The system can recognize how fast/slowvideo, where the edges are etc.
hc
|
