|
|
This legacy article remains on the web site for historical and research purposes. For a list of current article by Jan Ozer, please click here: |
THE CD-ROM PUBLISHER'S MPEG PRIMER
All rights reserved © Doceo Publishing, Inc. 1995
Need to get up to speed on the MPEG landscape in a hurry? This article presents a thorough overview of the current hardware and software players, and discusses how Windows 95 will help the MPEG installed base grow from under 150,000 to 20,000,000 in two years. This article is a must read if you're interested in publishing with MPEG, but especially if you're not interested in publishing with MPEG. First published in CD-ROM Professional, June 1995.
MPEG. This short form of "Motion Picture Expert Group" stands a solid chance of being voted the video acronym most likely to be found near a hyperbolic phrase or adjective in 1994 and 1995. Promoters label the technology a "steam roller" and predict an "MPEG" or "White Book" Christmas. Equally vocal opponents vilify MPEG, calling it an "obsolete technology." Some disparage MPEG as "designed by committee"; some predict this video compression/decompression standard will "die in committee."
On a totally objective basis, MPEG's growth on the desktop has been underwhelming. Mark Gaare, a senior analyst with market researcher In-Stat, Inc, estimates that only 200,000 MPEG decoders were sold by the end of 1994, with much of those numbers languishing on storefront shelves or warehouses. MPEG derivative CD-i has done well in Europe and Japan, but not in the US.
However, in this instance, what's past is clearly old news. Today, there are numerous CD-ROM publishers, video card manufacturers, codec developers and industry analysts, for whom MPEG's penetration into the CD-ROM world seems, in the words of one analyst, "as inevitable as death and taxes."
As with all technology shifts, timing and implementation are critical to successfully navigating through the changes to come. Market share will be won or lost, and opportunities exploited or frittered away. It is too early for title developers to throw away the QuickTime disks and Intel Smart Video Recorder, but it is high time for developers and publishers to start thinking about where and when to use MPEG.
But what are the various markets for MPEG technologies, and which MPEG-related CD-ROM products are bound for consumption on the PC? And what do the answers to these questions matter? A good place to start is with a look at MPEG itself.
The Motion Picture Experts Group -- MPEG -- is a joint committee of the International Organization for Standardization (ISO) and the International Electrotechnical Commission (IEG). The first MPEG Standard, known as MPEG-1, was introduced by this committee in 1991. Both video and audio standards were set, with the video standard built around the Standard Image Format (SIF) of 352x240 at 30 frame per second. MPEG data rates are variable, although MPEG-1 was designed to provide VHS video quality at a data rate of 1.2 megabits per second, or 150 KB/sec.
By resolution and data rate, MPEG-1 is targeted primarily at the computer and games markets. In contrast, MPEG-2, adopted in the Spring of 1994, is a broadcast standard specifying 720x480 playback at 60 fields per second at data rates ranging from 500 KB/sec to over 2 Megabytes (MB) per second. At these rates, MPEG-2 is not particularly relevant to the computer industry in the short term. However, MPEG-2 is expected by many to be anointed as the next digital broadcast standard for cable and network television.
MPEG-1 breaks into two general categories: those applications destined primarily for computer-based playback, and those bound for dedicated players like CD-I, MPEG-equipped 3DO, or the just emerging "Video CD" players. While there is some overlap, most developers focus on one or the other market. [Subsidiary MPEG-1 standards like White Book, CD-I, and Video CD were presented in the July/August 1994 issue CD-ROM Professional, in two articles: Mark Fritz's "Video CD, the Technology, and the Market: Dreaming of a White Book Christmas," and Matthew Leek's "MPEG Q&A" -- ed.]
MPEG: The Technology
MPEG is a codec -- short for encoder/decoder -- which means that MPEG technology is involved during both the compression and decompression of MPEG video. Unlike other codecs like Indeo and Cinepak, MPEG addresses both the audio and visual sides of video. Accordingly, each MPEG encoder must compress, and each decoder decompress both audio and video.
MPEG encoding divides into three technological categories; high-end software encoding, real-time capture and compression, and low-end or off-line software encoding.
High-end software encoding is performed by service bureaus like IBM Corporation, GTE Interactive Media, and Laser Pacific Media Corporation who use exotic high end capture systems to digitize the video and compress with million dollar supercomputers.
Next is real-time encoding performed with a board or board set from companies like Optibase, Inc., Optivision, Inc., and FutureTel, Inc, either purchased for in-house use or encoded by service bureaus using these products.
The final category, off-line software encoding, is the newest, and is currently occupied primarily by Xing Technologies Corporation and several capture board marketing partners. With the low-end, off-line approach, encoding is a two-step process, where first the video is captured, and then compressing with Xing's software encoder, Xing CD. Compression is performed totally in software and is extremely slow, even on a fast Pentium, where each second of video can need about one and one half minutes for compression.
In February 1995, Sigma Designs, Inc. introduced RealMagic Producer, a software/hardware combination which added a new wrinkle to low-end software solutions in the form of hardware-assisted compression. The RealMagic Producer cuts MPEG compression time down to about three seconds per one second of video, and represents an approach that can't help but make MPEG technology more accessible.
Whatever the approach, the common goal of all MPEG compression is to convert the equivalent of about 7.7 MB/sec of data down to under 150 KB/sec, which represents a compression ratio of about 52 to 1. While the 150 KB/sec rate is particularly relevant to CD-ROM developers, MPEG encoders can produce video streams both higher and lower than the data transfer rate of single-spin CD-ROM drives.
MPEG is a lossy compression technique, meaning that video quality drops as compression ratios increase: all things being equal, the lower the data rate, the lower the video quality. Despite marketing claims to the contrary, in most instances, MPEG video encoded at 150 KB/sec can't match the quality of the typical home VHS player. Depending upon the encoding technique, type of footage and analog medium, MPEG artifacts like blockiness, Gibbs Effect and mosquitoes are never totally missing, they're simply more or less apparent.
While MPEG is a very effective lossy compression technique, it does get stressed at 150 KB/sec. MPEG also isn't an absolute standard, but rather a specification for a compressed file format that can be created in many ways with varying degrees of success. All encoding schemes differ in quality and, predictably, the more expensive the capture and compression system, the better the quality.
Like most video compression schemes, MPEG uses both interframe and intraframe compression to achieve its target data rate. Interframe compression is compression achieved between frames, through, essentially, eliminating redundant interframe information. The classic case is the "talking head" shot such as with a news anchor, where the background remains stable and movement primarily relates to minor face and shoulder movements. Interframe compression techniques store the background information once, and then retain only the data required to describe the minor changes -- facial movements, for example -- occurring between the frames.
Intraframe compression is compression achieved by eliminating redundant information from within a frame, without reference to other video frames. MPEG uses the Discrete Cosign Transform algorithm, or DCT, as its intraframe compression engine. By and large, however, most of MPEG's power come from interframe, rather than intraframe compression.
MPEG uses three kinds of frames during the compression process: Intra or I frames, Predicted or P frames, and Bi-directional interpolated, or B frames. Most MPEG encoding schemes use a twelve- to fifteen-frame sequence called a group of pictures, or GOP.
I frames start every GOP, and serve as a reference for the first two B frames and first P frame. Since the quality of the entire GOP depends upon the quality of its initial I frame, compression is usually very limited in the I frame.
P frames refer back to the immediately preceding P or I frame, whichever is closer. For example, P frame 4 could refer back to I frame 1, and P frame 7 referring back to frame 4. During the encoding process, frame 4 searches frame 1 for redundancies, where the data about which are essentially discarded. Regions in frame 4 that have changed since frame 1 -- called "change regions" -- are compressed using MPEG's intraframe compression engine, DCT. This combination of interframe and intraframe compression typically generates a higher degree of compression than that achieved with I frames.
MPEG uses yet another compression strategy: B frames refer backwards and forwards to the immediately preceding or succeeding P or I frame. For example, for frame 11, a B frame, the compression scheme would search for redundant information in P frame 10 and the next I frame; once again, redundant information is discarded and change regions are compressed using DCT. The double-dose of interframe compression typically generates the highest compression of the three frame types.
All three types of encoders use the same basic GOP scheme defined in the MPEG specification. From a pure compression standpoint, the schemes differ in two key ways: their relative ability to identify interframe redundancies and whether they can modify GOP placement and order to maximize compressed video quality.
Although encoding details vary with approaches and particular tools, adherence to the MPEG file specification ensures that an MPEG video stream can be decompressed and displayed by all MPEG players. But there is plenty of variety when it comes to decoders, even if there are only three main types.
The first type of MPEG decoder is "overlay," where the decoder board connects to and displays through the computer's main video graphics adapter. Each overlay card contains dedicated MPEG video and audio chips that decompress the two MPEG components. Most early MPEG playback boards, including Sigma Design's original ReelMagic, are overlay cards.
The second category, which debuted in consumer quantities in late 1994, are "combo" boards, which combine both graphics and MPEG decompression functions. Rather than attaching to video cards, these boards replace video cards because they contain graphic accelerators chips along with the dedicated MPEG audio and video decompression hardware. Representative products include Matrox Electronic Systems, Ltd's Marvel II, Jazz Multimedia, Inc.'s Jakarta, and Sigma's RealMagic Rave.
Over 40 MPEG overlay and combo boards had been announced by the end of 1994. Overlay card street prices are now about $300, and are expected to break the sub-$200 barrier by end of year, 1995. Combo boards currently cost around $400, and should drop to under $300 within the next twelve months.
The third form of MPEG decode is software-only MPEG, which arrived in nascent form in mid-1994, but won't really mature until mid- to late-1995. Software-only MPEG decoders split the decompression task between the graphics card and host CPU, in order to decode MPEG audio and video without specific MPEG decode hardware assistance.
Only a few companies currently market software MPEG decoders, including Mediamatics, Inc, CompCore, Multimedia, Inc., and Xing. While strategies differ, these companies are primarily selling directly to graphics card manufacturers and computer manufacturers, rather than title developers or end users. Prices range from $2.50 to $5.00 per software decoder.
The low price of the software MPEG solutions can have a high price in performance today. Where both overlay and combo cards guarantee 30 fps playback, the performance for software-only MPEG decoders relates directly to the host computer. A 486/66 computer, for instance, currently produces about 8 fps playback, without audio playback. However, on a Pentium 90, all three of the major vendors of software MPEG claim to provide 30 fps of video with full audio support.
While finished software MPEG code is still very new and largely untested in the market, the first reports from technology demonstrations have been extremely positive. As a vendor of MPEG hardware playback cards, Jazz Multimedia's president, Jim Anderson, has every reason to discount software-only performance, yet he calls playback quality "surprisingly good," from CD-ROM on a Pentium 90 computer.
With today's computer/graphics combinations, software-only MPEG is more a novelty -- after all, much of MPEG's glamour relates not only to video quality but to the guaranteed 30 fps playback, which software-only solutions can't yet widely provide. In the long run, however, software-only MPEG may be the key to MPEG's mass market success. Indeed, the future seems fairly clear: MPEG will play back at 30 fps without dedicated decompression hardware, and the reduction in the incremental cost of MPEG playback could fall by about 99%.
NO MPEG IS AN ISLAND: THE COMPETING TECHNOLOGIES
But if the price of MPEG is on the way toward becoming, in effect, free across delivery platforms, will this video compression/decompression scheme be free of competition?
Indeed, not. There are other software-only codecs like Intel Corporation's Indeo and Radius, Inc's Cinepak which, like software-only MPEG,require no special hardware for decompression. In contrast to their "Free MPEG" counterparts, however, both Indeo and Cinepak perform well on a broader range of computers, producing 15 fps playback even on low-end 80486 computers.
The obvious benefits of software-only technologies to CD-ROM publisher and consumer alike are cost and flexibility. Decompression, which relies on the host CPU installed in every computer, is essentially free in the Indeo and Cinepak schemes, and this allows publishers to sell into a huge installed base of computers.
However, neither Indeo or Cinepak can touch MPEG when it comes to quality, and most CD-ROM producers using these schemes compress at 15 fps rather than the 30 fps, with resulting choppiness in higher motion video sequences. And at only 15 fps, pure frame quality is still noticeably worse than MPEG (at 30 fps); even at much higher data rates than that provided by standard speed CD-ROM drives, the quality per frame suffers, comparatively, with Indeo and Cinepak.
But the quality of these moving frames is a moving target, and the difference between MPEG quality and the software-only solutions is rapidly shrinking. For example, "talking head" videos, which are fairly easy to compress for all video technologies, are already similar in both Indeo (3.2) and MPEG formats. On Pentium/90 computers (keeping in mind that Indeo is a scalable codec, where the more powerful CPU provides the higher frame rate, and resolution results), an Indeo file will play at 30 frames per second at 320x240 resolution. Still, Indeo's progress is limited: on higher motion videos which stress all codecs, MPEG is the clear winner, even at close to half of Indeo's data rate.
Intel hopes to cut the quality difference even further with Indeo 4.0, now scheduled for release in August 1995. The new version will also focus on interactive features for CD-ROM developers, such as video hot spots that allow the user to click on areas in the video to create a response, and sprites, which are small video objects that can appear on top of graphics or other program elements.
Certainly, the backdrop of software-only technologies that improve with each new revision and general advancement in computer technology is an important consideration. Over the last twelve months, for instance, with the emergence of local bus as the video standard, specifications like the Display Control Interface (DCI), and the migration from 80486-based computers to Pentium-based machines have boosted software-only performance significantly. 1995 should also see Windows 95 brought to the market, with its 32-bit disk and CD-ROM drive access further enhancing video playback performance.
In contrast, MPEG -- even in its software-only potential -- is like Latin, an essentially dead specification that may be wonderfully functional today, but potentially obsolete tomorrow. Where Indeo 4.0 brings the promise of true interactivity, which is the holy grail of computer-based video, MPEG will deliver tomorrow the linear-play nirvana it delivers today -- 30 frames per second at native 352x240 resolution, scalable to full screen. MPEG may get cheaper, more integrated, and easier to work with, but it won't improve. On the other hand, MPEG proponents say, Indeo 4.0 is just that -- a promise -- while you can buy MPEG at your local supermarket, as it were.
But in the CD-ROM market, the difference between MPEG and software codecs are further exacerbated, because retrieving video data from a CD-ROM is more processor-intensive than retrieving data from a hard drive, especially at higher data rates. It can take twice as much processor power to retrieve data from a CD-ROM, which, in the zero sum game of processor-driven playback, robs CPU cycles from the decompression and display process.
| Transfer Rate | Hard drive | CD-ROM drive |
|---|---|---|
| 150 KB/sec | 11.1% | 21.1% |
| 300 KB/sec | 42.1% | 59.6% |
One practical consequence of the CPU load is that software-only video that may play flawlessly from a hard drive can stop and sputter when played from a CD-ROM. With hardware MPEG, where decompression and display is performed by the dedicated MPEG processor, hard drive and CD-ROM performance rarely differs. Of course, the price of this video quality and guaranteed display rate is MPEG decompression hardware; as software-only MPEG gains ground, it too will share the drawbacks of any other CPU-dependent decoder.
MPEG AND THE MARKET'S RESPONSE
Like many new technologies, MPEG suffered its share of growing pains. From the CD-ROM publisher's standpoint, the most significant MPEG roadblock may well have been with the application programming interfaces (which serve as the link between application programs and MPEG playback boards). Seeking to leverage its early lead into market dominance, Sigma Design kept their MPEG API proprietary, which meant that titles produced to the Sigma API couldn't play on other MPEG boards.
However, in June 1994, Microsoft released the MCI MPEG playback specification, a Windows-based playback API now used by all playback boards, including those manufactured by Sigma Designs. This new standard ensures that all properly developed Windows programs will play back on all MPEG playback devices.
The early API problem wasn't the only problem MPEG experienced. On the end-user side, for instance, technical limitations relating to the overlay design used by most MPEG playback boards caused significant installation and support woes, and, ultimately, many product returns. Early MPEG playback cards proved incompatible with many graphics adapters, especially (and ironically) the more advanced graphics cards used by early technology adopters most likely to purchase MPEG playback cards. In some instances, "leading edge" users were forced to reduce display color depth to 16 or even four colors, which had to be a unpleasant consequence to keep up with the most current technology, seeing that the net result was a step or two back to the CGA days of 1985.
Playback MPEG board manufacturers have been attacking these problems on two fronts. To improve overlay compatibility, MPEG manufacturers have invested in compatibility testing and additional software development to stabilize their drivers. More importantly, the introduction of combo boards which combine the graphics adapter and MPEG playback functions avoid the compatibility problem altogether.
Nonetheless, MPEG sales have not yet rebounded from the start up woes, and sales of MPEG decoder boards have languished while the multimedia market booms around it. In Stat's earlier referenced figures -- estimating that only about 200,000 MPEG decoder boards had shipped though the end of 1994 -- are supported by market research giant Dataquest's own informal numbers for MPEG.
The problem? MPEG's own great chicken and egg question, where end users won't buy MPEG products without titles to watch, and where title producers won't build titles without an installed base of end users.
At least in the short term, this MPEG roadblock shows little sign of relenting.
To date, few publishers -- if any -- have released MPEG titles without a bundling agreement with an MPEG playback card company or computer seller. A survey of the CD-ROM publishing field reveals a wide-range of attitudes -- with comments such as "Here's a quarter, call me when MPEG's a market," to "MPEG's coming and we're gonna be ready" -- but a hard-nosed practical mentality is a common theme.
Typical of the views of CD-ROM publishers is Tony Bove, president of California-based CD-ROM developer Rockument, whose first title, Haight Ashbury in the 60s, is being released using Cinepak for the substantial video content on the disc. Perhaps better known as the founding editor of the industry newsletter, The Inside Report on New Media, Bove states, "We don't need the best quality video, we need the best quality video that plays on computers owned by our customers. And today, few have MPEG."
Rafael Laderman, of Kozel Multimedia in New York City, the company that developed the CD-ROM JFK Assassination, a Visual Investigation, concurs. "We don't see MPEG in the short term," Laderman insists. "We focus on the installed base."
Intellimedia Sports, who develops games and instructional sport titles for the Macintosh, Windows, and 3DO platforms, is more bullish. Intellimedia uses Cinepak in every title but one -- Cowboy Casino -- which is bundled with Sigma's Real Magic.
From his office in Atlanta, Intellimedia Sports vice president of Development, Bill Markle, comments, "There's some possibility that MPEG will appear on computer motherboards by Christmas [1995], and we can move in a hurry if this happens. While we don't simultaneously produce titles in both Cinepak and MPEG, it would take less than 6 weeks, and cost less than $12,000 to convert from Cinepak to MPEG."
In a similar vein, Dov Jacobson, VP of Creative Development at Atlanta-based Turner Interactive, producer of the Gettysburg CD-ROM, comments that his company doesn't release MPEG titles without a bundling agreement. "We're presently discussing several opportunities, however" Jacobson remarks, "We consider all titles with high video content MPEG candidates, and develop from the start with MPEG in mind. The cost of converting to MPEG is insignificant."
Nonetheless, none of the mainstream title publishers have adopted MPEG full-bore, and no title developers have publicly announced any comprehensive MPEG strategies. This bodes poorly for the MPEG market in the short term.
In truth, the hardware-assisted MPEG market may never take off. So says Dataquest analyst, Kathy Klotz, who predicts a very short, interim market for both overlay and combo boards. "With the power of the Pentium," Klotz states, "you've got the capacity for high quality software-only video which meets the demands of most end users. We see a very small market for MPEG decoder boards."
Indeed, if one were to adopt the historically supported "software sells hardware" viewpoint, MPEG's future would be bleak. So why are most industry analysts bullish on MPEG? Because, as Richard Doherty, founder of Envisioneering Inc., a marketing research firm in Seaford, NY, predicts, "Several market factors are combining to produce over 20 million computers with some form of MPEG playback by the end of 1995."
THE REAL MPEG MARKET: BORN IN A BUNDLE
How is MPEG playback going to end up on over 20 million computers by the end of 1995? The answer lies with the graphic chip manufacturers and graphics card manufacturers aligning themselves with one of the three software-MPEG providers, to bundle MPEG software with their chips and graphics cards.
Typical of the emerging MPEG bundles is the agreement between Western Digital and Mediamatics concerning Western Digital's recently announced the Rocketchip WD974. Western Digital announced the agreement in their press release in the early months of 1995, along with claims that "systems based on [the Mediamatics technology] will deliver cost-effective motion video at 30 frames per second, at prices lower than solutions on the market today." Envisoneering's Doherty reports that chip makers Brooktree, Tseng Labs, S3, and all other major graphics chip vendors have either signed MPEG-related agreements or are in the final stages of negotiation.
Graphics card manufacturers confirms Doherty's comments. Canadian manufacturer Matrox Electronic Systems Ltd, had only one MPEG playback device -- the combo card Marvel II -- on the market at the end of 1994, and this was focused towards OEM, rather than consumer channels. However, while they haven't announced any specific products, Matrox is clearly looking to expand both hardware and software MPEG into their consumer product line. Caroline De Bie, Matrox's media relations representative for the US, comments, "We believe that hardware MPEG will become a standard for high-end graphics cards and multimedia systems by the end of 1995 for high end multimedia systems. We see software MPEG as a good solution for mainstream end users, especially game users who see MPEG as an entertainment."
Phil Parker, from graphics card manufacturer Number Nine Computer Corporation, echoes De Bie's comments. "We believe that MPEG playback -- in either hardware or software -- will be a necessary feature for graphics cards," Parker says. "We're developing accordingly."
Parker indicates that Number Nine will support software MPEG in their low- and mid-level lines and hardware MPEG on high-end products. In fact, the company seems to see MPEG as an opportunity to capture market share in low- and mid-range graphics markets. "Some people think that Number Nine only makes high-end graphics cards," Parker states. "This year we're going to change that perception." Parker predicts that some form of MPEG would appear on two-thirds of his company's products by year end 1995, which is quite a jump considering that Number Nine started the year without any MPEG entries.
International Data Corporation analyst Louise Biggs thinks that computer vendors will take the matter into their own hands if graphics card/software MPEG bundles don't arrive before the Christmas buying season. "Computer makers can spend $2.50 on the Mediamatics player, call themselves an MPEG playback station, and differentiate their product from the competition," she says. "It's a no brainer."
A number of analysts see Windows 95 boosting software MPEG penetration. "The trend is unmistakable," Envisioneering's Doherty says. "Although Microsoft is not including MPEG software playback in Windows 95, it is giving their full support to MPEG."
"When Windows 95 ships," Doherty continues, "Microsoft will also introduce a companion disk containing a series of Windows 95 companion products. Pricing isn't set, but it should be under $99, and perhaps as low as $19.95. That CD will contain software MPEG from Mediamatics."
Doherty believes that most users who upgrade to Windows 95 will purchase the disk, and he suggests that CD-ROM publishers should consider all computers with Windows 95 to be MPEG playback stations in their market planning. Between software MPEG players shipping with graphics cards, and those purchased with Windows 95, he predicts that 10 to 20 million computers will be equipped with software MPEG by the end of 1995.
IDC's Biggs predicts that the swelling of the installed base of MPEG will have a significant effect on CD-ROM publishers. "Today, Indeo has hearts and minds of title developers, but graphics card and computer manufacturers vendors drive the market and all the momentum is behind MPEG, not Indeo," says Biggs. "Title developers aren't loyal -- their business model dictates following the largest installed base, and graphics and computer vendors create the installed base."
Doherty agrees with Biggs, and predicts that 25% of titles shipping by Christmas 1996, will be MPEG titles, up from an estimated 5% for Christmas 1995.
THE CPU AS THE GREAT LEVELER: SOFTWARE MPEG VERSUS INDEO AND CINEPAK
Even the mid- to long-term picture for "software" MPEG looks plenty rosy, such forecasts don't mean that software codecs like Indeo and Cinepak are going away any time soon. First, there will remain an estimated 60 to 80 million 486/Pentium computers that will never be powerful enough to run software-only MPEG, and only few of these machines are ever expected to upgrade to hardware MPEG. IDC's analyst Biggs admits that software-only MPEG is a "forward looking solution that does little for the installed base of computers not powerful enough for software MPEG."
But betting on markets that disappear when faster silicon arrives is a perpetually bad investment. In truth, the early software codecs succeeded primarily because they were the only game in town -- no one was really enamored with the quality of video produced, even with the excitement and opportunity brought to bear by QuickTime and Video for Windows.
As a group of software codecs, Indeo, TrueMotionS, and Cinepak spent 1994 eking out marginal improvements, while some really bright engineers at CompCore, Mediamatics, and Xing were making 30 fps, software-only MPEG possible on an increasingly relevant class of computers. Everyone knew that software-only MPEG was possible. The surprise has been that software MPEG is becoming as good as it is as fast as it is.
However, 1995 is bringing dramatic advances in other software codec technologies, including expectations for an advanced fractal codec from Iterated Systems, further advances in TrueMotionS from partners Horizons Technology and Duck Corporation, and the eagerly awaited Intel's Indeo 4.0; even Radius is said to have aggressive plans for Cinepak. But the key advantage held by these codecs, which has been the ability to run without hardware, is disappearing.
And there are a range of advantages coming forward behind MPEG. One circumstance in MPEG's favor is a stable crop of high-end companies like Optibase and FutureTel that are helping to anchor MPEG technology. There is an increased competition for decoder chips -- from such companies as C-Cube, Motorola, S3, and Zoran -- that are dropping the price of dedicated MPEG playback, and significantly decreasing encoding costs are predicted to drop as competitive silicon enters the market. Perhaps most importantly, however, is the general acceptance of MPEG as the emerging compression standard for broadcast and cable, because these two markets will increasingly intersect with the computer over the next few years.
These factors make it very difficult to imagine any CD-ROM market scenario where MPEG won't be pervasive on future generations of computers: if software-only MPEG doesn't produce 30 fps on the current crop of Pentium 90 computers, it will on the next generation of Pentiums and its successor P6. But the rule of MPEG seems likely even on many current platforms, because, as Evisioneering's Richard Doherty points out, "Software-MPEG doesn't have to be as good as hardware MPEG -- it just has to be better than Indeo." Like most other changes in areas where installed base is built one machine at a time, the shift toward MPEG will not be immediately dramatic, but rather realized with increasingly swift pace of the next eighteen months.
Smart publishers are already planning for MPEG. Compact Publishing/Softkey International's president Robert Ellis anticipates that most of their 1995 multimedia titles will remain Indeo 3.2, software-only video, but he sees a clear trend toward the adoption of MPEG by their OEM customers. "We plan to introduce at least two MPEG titles in 1995," Ellis remarks, "one of which is the 1995 TIME Almanac."
Indeed, CD-ROM publishers that haven't already started working with MPEG, may be missing opportunities, especially as MPEG bundling increases dramatically over the next twelve months as MPEG gets added to graphics cards and computers. As Turner Interactive's Dov Jacobson says, "We'd be foolish not to pursue large scale bundling opportunities that can move 50,000 units into a market that doesn't compete with our base, software-only business."
Jan Ozer is president of Doceo Publishing, a multimedia title developer and publisher of the Video Compression Sampler series. A frequent contributor to CD-ROM Professional, Ozer is author of Video Compression for Multimedia, published by AP Professional in late 1994. Ozer can be reached at Doceo Publishing, Inc., 568 14th Street, NW, Suite 200, Atlanta, GA 30318, at (404) 892-2889, or jan@doceo.com.
Companies Mentioned In This Article
Advanced Digital Systems 13909 Bettencourt Street Cerritos, CA 90703 310/926-1928 Fax 310/926-0518 Brooktree Corporation 9868 Scranton Road San Diego, CA 92121-3707 619/452-7580 Fax 619/452-1249 Compact Publishing/Softkey International 201 Broadway Cambridge, MA 02139 617/494-1200 CompCore Multimedia, Inc. 1270 Oakmead Parkway, Suite 214 Sunnyvale, CA 94086 408/773-8310 DataQuest, Inc. 1290 Ridder Park Drive San Jose, CA 95131 408/437-8000 Envisioneering, Inc. 3864 Bayberry Lane Seaford, NY 11783 516/783-6244 FutureTel, Inc. 1092 East Arques Avenue Sunnyvale, CA 94086 408/522-1400 GTE Interactive Media 2385 Caminor Vada Roble Carlsbad, CA 92009 619/431-8801 IBM Corporation Old Orchard Road Armonk, NY 10504 800/426-3333 In-Stat, Inc. 7418 East Helm Drive Scottsdale, AZ 85260-2418 602/483-4464 Intel Corporation 5200 N.E. Elam Young Parkway Hillsboro, OR 97124-6497 800/538-3373 Intellimedia Sports, Inc. Two Piedmont Center, Ste. 300 Atlanta, GA 30305 800/269-2101 International Data Corporation Five Speen Street Framingham, MA 01701 508/872-8200 Jazz Multimedia, Inc. 1040 Richard Avenue Santa Clara, CA 95050 408/727-8900 Kozell Multimedia 380 Lafayette Street New York, NY 10003 212/533-2650 Laser Pacific Media Corporation 809 North Cahuenga Boulevard Hollywood, CA 90038 213/462-6266 Matrox Electronic Systems, Ltd. 1055 St. Regis Boulevard Dorval, QC H9P 2T4, Canada 514/685-7230 Mediamatics, Inc. 4633 Old Ironsides Drive, Ste 328 Santa Clara, CA 95054 408/496-6360 Number Nine Computer Corporation 18 Hartwell Avenue Lexington, MA 02173 800-GET-NINE Optibase, Inc. 5000 Quorum Drive, Ste 700 Dallas, TX 75240 214/774-3832 Optivision, Inc. 4009 Miranda Avenue Palo Alto, CA 94304 415/855-0200 Radius, Inc. 215 Moffett Park Drive Sunnyvale, CA 94089-1374 800/227-2795 Rockument P. O. Box 1289 Gualala, CA 95445 707/884-4413 Sigma Designs, Inc. 46501 Landing Parkway Fremont, CA 94538 800/845-8086 Smart & Friendly 16539 Saticoy Street P.O. Box 9277 Van Nuys, CA 91409-9277 818/994-8001 Fax 818/988-6581 Turner Interactive 1050 Techwood Drive Atlanta, GA 30318 404/885-4885 Xing Technology Corporation 1540 West Branch Street Arroyo Grande, CA 93420 805/473-0145 Western Digital Corporation 8105 Irvine Center Drive Irvine, CA 92718 800/275-4932
MPEG. This short form of "Motion Picture Expert Group" stands a solid chance of being voted the video acronym most likely to be found near a hyperbolic phrase or adjective in 1994 and 1995.
MPEG's growth on the desktop has been underwhelming. . . However, in this instance, what's past is clearly old news.
By resolution and data rate, MPEG-1 is targeted primarily at the computer and games markets.
In February 1995, Sigma Designs, Inc. introduced RealMagic Producer, a software/hardware combination which added a new wrinkle to low-end software solutions in the form of hardware-assisted compression.
Over 40 MPEG overlay and combo boards had been announced by the end of 1994.
But if the price of MPEG is on the way toward becoming, in effect, free across delivery platforms, will this video compression/decompression scheme be free of competition?
However, neither Indeo or Cinepak can touch MPEG when it comes to quality, and most CD-ROM producers using these schemes compress at 15 fps rather than the 30 fps, with resulting choppiness in higher motion video sequences.
The early API problem wasn't the only problem MPEG experienced.
To date, few publishers -- if any -- have released MPEG titles without a bundling agreement with an MPEG playback card company or computer seller.
Tony Bove states, "We don't need the best quality video, we need the best quality video that plays on computers owned by our customers. And today, few have MPEG."
How is MPEG playback going to end up on over 20 million computers by the end of 1995? The answer lies with the graphic chip manufacturers and graphics card manufacturers aligning themselves with one of the three software-MPEG providers. . .
IDC's Biggs predicts that the swelling of the installed base of MPEG will have a significant effect on CD-ROM publishers.
If software-only MPEG doesn't produce 30 fps on the current crop of Pentium 90 computers, it will on the next generation of Pentiums and its successor P6.
Evisioneering's Richard Doherty points out, "Software-MPEG doesn't have to be as good as hardware MPEG -- it just has to be better than Indeo."
[Sidebar Ozer]
MPEG has gained a reputation as the high quality video compression scheme. And quite appropropriately, if not, however, without cost for today's computer platform. Several hundred dollars worth of MPEG decode chips are required for top-quality playback.
But just because hardware-assisted MPEG wins in the resolution and frame rate race against the software-only Indeos and Cinepaks of the world, that doesn't mean that MPEG video is faultless. Despite wide reports, most MPEG video encoded at the CD-ROM target data rate of 150 KB/sec can't match the quality of VHS, although depending on how dirty a VCR one has, or the generation of analog tape being viewed, the common comments favorably comparing MPEG and VHS can seem accurate enough.
But in fact, like any lossy scheme, MPEG has artifacts. Among the most common categories of MPEG compression artifacts are Gibbs effects, blockiness, and "mosquitoes."
 | Gibbs effects, one of the most common MPEG (and JPEG) artifacts, show up most in high contrast border areas, such as around letters or numbers. |
 | Blockiness in MPEG is actually an artifact of the JPEG component in the scheme, and it shows up during extremely high motion video sequences. The blockiness reflects the basic JPEG characteristic of dividing each frame into 8x8 blocks (quantizing) and compressing each of those blocks separately. |
 | While Gibbs effects are specifically associated with artificial images like the geometric forms of text or numbers, "mosquitoes" are a similar effect that can occur around natural objects with the result that areas of the video can shimmer with the motion of the objects. |
-- Jan Ozer