Editable MPEG-2......................................
tvbroadcast.com
Is MPEG-2 Edit-Friendly?
By Bob Turner
At this year's NAB, Sony's "Digital Reality" featured "MPEG World" as part of its convention theme. Spokesmen evangelized Sony MPEG editors as the next generation of products and technologies that enable an all-digital infrastructure.
Sony's MPEG strategy includes five key "advantages:" Selectable picture format, MPEG is scaleable from SDTV to HDTV (1080p, 1080i, 720p, 480p, and 480i); selectable picture quality, bit rate scalability up to 300 Mbps; high-quality with a low data rate (with long GOP), the reason it was selected as the broadcast distribution format; editability, I-frame only or short GOP; and flexibility, users can select a data rate--making it suitable for multiple network architectures.
As stations move toward a digital future, many (including CNN) agree with Sony's vision for the MPEG compression scheme.
"We see MPEG-2 video as the way of the future," said John Davis, president and CEO of Applied Digital (a manufacturer of a family of ADedit MPEG-based editing products. "Certainly it offers satellite transmission and storage requirement benefits with which JPEG and uncompressed video cannot compete. MPEG-2 already has become the standard for broadcast, transmission and commercial ad insertion. And HDTV is MPEG-2 based. There is little doubt that more and more broadcast activities are going to involve MPEG-2 compression. It stands to reason that MPEG-2 editing is a most desirable element of the MPEG-2 snowball."
It would seem logical that remaining in the MPEG-2 environment would allow users to edit content that is originally MPEG-2 and keep it in that format without any degradation of video quality.
"Why go through the trouble of decompression and recompression or video transcoding to perform an edit just to take it back to MPEG again?," he added.
Yet, people like Phil Livingston, vice president of Engineering Support and Training, at Panasonic Broadcast and Digital Systems offer a conflicting view. "MPEG proponents would have you believe that the world will soon have a roster of harmonious MPEG-compliant choices, and choices of signal formats and interoperable equipment built upon them," he said.
"Panasonic understands this expectation, but we suggest that the roster may well be neither seamless nor ubiquitous--and for both economic and technical reasons it will not be dominated solely by MPEG."
It's interesting to note that Avid Technologies, which showed one of the first prototype video engines for editing MPEG video at NAB '96, now offers several uncompressed editing solutions (including its newest Media Composer, shown at ITS in Los Angeles and IBC, in Amsterdam) but does not offer an MPEG-based editing system.
"Avid is not at all opposed to MPEG", said David Schleifer, broadcast marketing manager of the Broadcast News Division at Avid. "We are quite active in this technology; we continue to explore the problems of editing."
Yet, he continued, "Avid Technology is not a compression format company, and as a result we are capable of supporting any format the market commits to."
MPEG-2 Editing
"MPEG-2 is the clear future of digital video," offered Richard Doll, senior product manager at Fast Multimedia US. Fast markets its 601 system, an MPEG-based nonlinear editing solution. "Only MPEG-2 and DV/DVCPRO have been accepted by the joint SMPTE/EBU taskforce."
MPEG was designed as a distribution format using asymmetrical processing. This involves a more sophisticated (and expensive) encoder with a simpler (and less expensive) decoder. While no one doubts its value for transmission, the question on many professionals' minds is whether this technology best suited for production?
"The fundamental concept of point-to-multipoint dictated that one of the foundations would be inexpensive decoding at the expense of complex encoding, "Livingston said. "For example, a realtime DTV encoder [that will be used by TV stations for transmission] can cost hundreds of thousands of dollars, while the set-top box should cost a few hundred dollars and use a decoder that costs tens of dollars. This concept is of little value in production equipment, where one can easily imagine that each piece contains both encoder and decoder."
Uncompress/recompress Generational Loss
But is MPEG compression lossless? The fact is that whenever an editor creates a dissolve, key, DVE move or process the signal in any way, they must uncompress and then recompress the material.
"Even Fast cannot beat mathematics," Doll admitted. "You simply cannot perform image manipulations on a pixel level (and this is what all these effects and DVEs are) on a compressed basis. Every compressed image must be decompressed to render or display these manipulations. The digital generation loss is indeed an issue here. But as a typical NLE system does perform all render processes for a particular frame in a single step, only one recompression step will be performed, no matter how many layers of effects are involved.
"Independent tests by such organizations as the SMPTE, EBU and IRT (Institut fuer Rundfunktechnik) have proven that MPEG-2 422P (I) at 50 Mbps is visually transparent for up to seven generations," Doll added. "This provides image quality nearly as good as Digital Betacam and by far more than enough for professional quality nonlinear editing. Also remember that simple cuts do not require any recompression."
"Producing polished, locally-branded product from the network's MPEG-2 transmission requires conversion from MPEG-2 into base-band video and audio,", said Fred Schultz, product manager for ASC Audio Video Corp., a Leitch Company. "Once in base-band, transitions can be performed on any frame and effects can be multilayered. Keying, including logos and crawls, become possible and keys can be transparent."
"Because no original MPEG profile was suited to professional studio applications, a new MPEG 4:2:2 Profile (the so-called "Studio Profile") has been created, allowing standard definition rates up to 50 Mbps, depending upon the complexity of the images," added Panasonic's Livingston. "Industry groups are now discussing establishment of standards for 100 to 150 Mbps and 300 Mbps profile plateaus with minimal GOP structure for HD."
Schleifer stated, "There are many technical issues relating to MPEG that are being resolved in the proper standards bodies, and Avid is actively participating in the discussions. Some of the major issues to be tackled are that MPEG allows for differences in encoding as long as playback is consistent, and that MPEG variants with large GOPs are not well suited for frame accurate rolling and editing."
Since vendors design their systems' encoding parameters to match their application (editing, streaming, CD-ROM, etc.), the vision of consistent MPEG encoding--from acquisition to consumer--has not been realized. As conventions arise surrounding the use of MPEG for editing, like I-frame only MPEG, using them will need to be reconciled against the need to use other very different forms of MPEG in the broadcast process.
"The fundamental MPEG compression efficiency, which is based on longer GOPs, is lost in the post production process," observed Livingston. "The MPEG-2 standard does not in itself provide for transcoding or transformation between profiles --rather the standard requires compliant decoding, where in any stream of equal of lesser syntax (lower profile level) must be decoded."
Cascading Compression
"Cascading compression schemes are not a real good thing for digital video," Doll said. "The lower the initial quality, the bigger the quality loss. But this loss is even worse if you are using different compression scenarios, such as going from DV to M-JPEG to MPEG-2 MP@ML. In addition, MPEG-2 offers excellent quality at lower data rates than a comparable M-JPEG codec."
Quantel's Pannaman said, "MPEG-2 could have advantages when compressing an I-frame-only video stream to the MPEG-2 transmission format, but in truth that is not proven. It seems to be a logical conclusion, but so have technological concerns in the past that have proven to be invalid."
Frame Accuracy
"There have been discussions about the frame accuracy of edits made with MPEG-2 edit systems--especially when streaming the output at 4X speed," Doll stated. "If you work with 422P@ML I-frame-only signals, you have very little reason to worry about frame accuracy of MPEG editing systems,"
Doll added, "Once you are in an MPEG-2 IPB stream, you probably don't want to go back to your edit desk. All you need to do at this late stage is to do what is known as 'seamless splicing,' that is, a seamless insertion of other footage in an existing distribution stream. This splicing does not necessarily have to be frame-accurate, although this can be performed with very little quality loss as the broken GOPs around the splice has to be re-compressed."
"MPEG is by definition motion compression," added Applied Digital's Davis. "Inherent to the temporal compression scheme of MPEG are difficulties in frame-accurate editing. However, [we've] developed proprietary techniques whereby our ADedit MPEG-2 Editors maintain the frame accuracy and GOP integrity that other MPEG-2 editing systems cannot."
Interoperability
"One of the [disadvantages] of MPEG-2 editing is interoperability." admitted Doll. "However, the MPEG-2 Taskforce is making attempts to solve these problems; enabling MPEG devices from different vendors, be they video or computer, to work together seamlessly and in native digital quality."
MPEG Vs. DVC
"DV compression is now widely accepted--especially for broadcast news," said Pannaman. "MPEG-2, meanwhile, appears behind in the level of development."
Livingston points out that, "The consortium of companies that developed the DV compression engine recognized both the power of MPEG and the limitations as well, because MPEG compression already existed."
"There are, however, situations where MPEG editing makes more sense," said Doll. "MPEG offers full chroma bandwidth and high data rates, making it more suitable for color keying and/or multilayering and compositing than DV. Also, as DV compresses the outer parts of a picture with higher compression ratios than the middle, lower-third titles might be a problem in DV-based post production. Editing with MPEG solves these restrictions, although the price you pay is [related to] disk space--due to the lower compression factor."
Conclusions
According to Quantel's Pannaman, it appears that the greatest benefit of MPEG comes when a facility stays in a native compression system throughout the plant. "The acquisition tape format selected will be a major criteria for selecting a compression method for your nonlinear editing system."
Schleifer stated, "There really are no broadcast news nonlinear MPEG editors shipping in the marketplace today. All of the systems that are available are struggling with issues that need to be resolved by the proper MPEG standards bodies. Until these issues are resolved, any MPEG editing implementation will either be delivered with compromises, or with a proprietary or non-standardized implementation.
"We strongly believe that the benefits of a nonlinear editor," he concluded, "or of a system and workflow built around such systems, will come from the features it supports and benefits it delivers. As long as the compression format delivers or exceeds the quality required, this type of solution will find a home with broadcasters."
|
