Compression systems............................
broadcastengineering.com
In terms of shared files, a major development in the field of compression in general is C-Cube's release of its DVxpress architecture, which will allow a new generation of editing products to work in a mixed DV/MPEG environment. This is a great new idea, and all the products involved certainly are going to get more efficient in the year to come.
Broadcast Engineering June 1999
Compression systems
By Mark Boeddeker
The term video compression stands for many things these days, and every one of them was represented at this year's NAB. Basically, video compression describes the various ways in which redundant information is removed from a digital video stream so humans do not notice the missing information. The various tape-based digital acquisition formats on the market (S-Digital, DV, DVC, Digital Betacam and SX) all use a compression scheme of some kind, as do many satellite uplinks, cable on demand or DSS systems. To determine acceptable visual quality levels, human perceptual models have always been used, but as DTV and HD implementation proceeds, the quality model should become what is better and best. Ultimately, different eyes will see different things; therefore beauty, and perceived digital artifacts, will continue to be in the eye of the beholder.
MPEG-1 was designed to transfer data from a T1 line or single-speed CD-ROM. It has been widely implemented in all manner of applications from games and CD-ROM to video on demand and point of purchase, but broadcast applications are limited. At video CD data rates, it aspired to VHS quality in a digital format. At slightly higher data rates (in the 3-4Mb/s range), MPEG-1 looks comparable to MPEG-2 at the same data rate, especially with line-doubling or other buffered techniques used at the set-top. MPEG-2 was designed to support a much wider range of resolutions and data rates. M-JPEG compression has produced good-looking imagery for years and has been the compression scheme of choice for NLEs. Many applications will continue to use M-JPEG, even though MPEG has become the standard for broadcast and DTV delivery. One reason is that MPEG-2 delivers a broadcast-quality image at approximately 50 percent of the data requirements of M-JPEG. Editing MPEG is still problematic, but products on the market already claim to address most problems.
As analog tape formats begin to fade away in the broadcast world, MPEG servers will take their place. Feeding those servers will take a myriad of devices, leaving quite a bit of flexibility as to exactly how it's done. Integrating new technology is challenging, but systems and hardware exist to answer any operation's needs.
Sony's SX product line is a system of native MPEG-2 acquisition, editing and transmission products that no one really knew what to do with a few years ago. Things have changed. Staying MPEG throughout the acquisition to the transmission process has decided advantages. Several of the networks have implemented SX products in their operations, and it is one of the less painful routes to DTV and HD. The media-server technology that supports SX is an example of how multiple workstations can access the same digital image base.
Integrating multiple workstations and formats on one shared database requires file compatibility on a major level. Often one would choose to keep certain investments working (even if they do generate obsolete file structures). Products from companies like Centra-Vision/MountainGate create file systems that allow products running on different operating systems to access the same database in their native file formats.
In terms of shared files, a major development in the field of compression in general is C-Cube's release of its DVxpress architecture, which will allow a new generation of editing products to work in a mixed DV/MPEG environment. This is a great new idea, and all the products involved certainly are going to get more efficient in the year to come.
Although encoders exist for most computer operating systems, Mac-based encoders led the way in creating files for the early MPEG-1 applications. Mac-based Minerva and Sonic Solutions products proved very compatible with DVD applications early on, but PC-based encoders from companies like Optibase and Digital Vision are keeping pace. The simplicity of Linux would lead one to expect more products running on that OS soon.
One thing to keep in mind is that the MPEG specification is a multifaceted gem that includes simple, main, and high Profile specs at low, main and high Levels, as well as several others. For example, MP@ML refers to a main profile (4:2:0 quality) at main level (601 resolution) MPEG environment. MP@HL refers to 4:2:0 quality at high level or HD resolution.
What this means is that there are many types of MPEG, and you should be sure you know what kind you need. Higher-priced encoding products in the $100,000+ range are capable of customizing the encode parameters of an MPEG stream, both audio and video. This makes them capable of delivering virtually any type or quality of MPEG file. If you need everything, you'll have to pay for everything.
There are several capable encoders in the "less than $100K" category that are often seen bundled with DVD systems. Other encoders are only available as part of a DVD authoring product. If you intend to do DVD exclusively, some of these minimum configuration encoders may be sufficient, but more flexibility in the encoder will let you use it for other things. Depending on your video source, pre-processing may be required so you are not wasting bits trying to compress noise. Several pre-processors were on display this year including the Prefix from Snell & Wilcox (see Pick Hits). Digital Vision introduced the BitPack-HD, an HD offline or premastering system for video servers and DVD. The BitPack-HD creates HD files for 1080i/720p formats using the same BitPack control environment as standard definition.
For encoding, a common feature is the ability to do CBR (constant bit rate) as well as VBR (variable bit rate) encodes. VBR encoders use an average bit rate process that allocates more data to be used in visually complex or problem areas. Some products utilize multiple-pass encoding which produces a finer analysis of the video in terms of things like scene change detection, theoretically resulting in a better encode. Options such as video preprocessing, inverse telecine detection and the ability to re-encode problematic video segments are other features that can add to the final price of any encoder.
MPEG audio is standard with most encoders, but, for many applications, PCM or Dolby AC-3 is required. This is an option you pay for most of the time. Integrated audio encoders work just fine and sometimes have an advantage in terms of maintaining sync with the associated video, but an audio-only facility has some options when it comes to stand-alone PCM and AC-3 encoders.
A new generation of low-cost encoder/decoder products is also entering the marketplace. For dedicated applications, some of these products may be completely adequate. One thing to consider is that the single board, single chip type of encoder is not going to have the flexibility to generate the more esoteric MPEG streams, and artifacts are more apparent in a low data rate CBR type of encode. Decoders need to be able to handle high data rates without choking or locking up.
MPEG decoders (play devices) usually require a multiplexed or muxed file for playback. This is a file in which the discrete video files (.mpeg, .mpv) and audio files (.ac3 or .mpa) are merged into one file. Encoders can do this on the fly in real time, or they can deliver elemental (discrete) audio and video files which are muxed later. Some end use applications require special recipe multiplexing, which is usually done after the initial encode with a dedicated multiplexing tool, such as the applications available from Pixel Tools, Heuris or Manzanita Systems.
Elemental MPEG-2 audio and video files are multiplexed into transport or program streams. Transport streams are designed for use in potentially high error environments, such as transmission. They include timing elements that make them more error tolerant and so, less dependent on specific playback parameters. Program streams function in a variety of software environments and are less error tolerant. Many applications and/or hardware devices will only play one stream type or the other. For those wanting to analyze these streams, Interra showed its MProbe MPEG bitstream analysis and compliance software which can be very helpful when looking for problems in these areas.
Software encoder/transcoders can give new life to many systems by adding the ability to create MPEG-2- and DVD-compliant streams from a variety of sources, including DV and .avi files. Some products include a hardware assist in this conversion, but even with dual processors, software crunches are time consuming.
For now, we will continue to get what we pay for. Many end use applications don't require the highest quality MPEG files available, and can be supported by one of the less expensive encode/decode solutions. On the other hand, the expensive encoders are more flexible in terms of file type and encode parameters, and can produce higher-quality, higher-bandwidth encoding than is possible with the cheaper gear.
Among the many products shown at this year's show, several were interesting including OpenTV's OpenAuthor 2.0 software authoring tool that is designed to help create interactive television programming. The software includes 32 built-in gadgets and allows them to import and preview content. STB Systems (now 3dfx) showed its Quartet series of multiport MPEG decoder cards, capable of decoding four MPEG streams using a single PCI slot. For those looking to make multiple CDs on a small scale, Telex showed the CDP 2001, a desktop CD duplicator that supports all standard writing modes and is easily expandable to support DVD-R and CD-R.
Finally, ISLIP Media demonstrated the MediaSite AutoLogger and MediaSite BuilderNT. The Autologger allows users to encode, catalog and index video of browsing and retrieval over intranet or Internet. MediaSite BuilderNT analyzes and indexes video based upon Autologger metadata and the video information itself.
|
