Related Articles

Faced with DVD Video's increasing popularity, we shooters would do well to become more familiar with the requirements of MPEG-2 encoding. Apple's new DVD Studio Pro and similar products are opening a whole new world to many of us hoping to produce our own DVD titles on the desktop.

As more and more shooters and image professionals are discovering, however, the move to DVD is not without its challenges. Despite the format's touted image quality, the fact remains that MPEG-2 compression entails substantial compromises. These often fly in the face of the shooter's hopes to maintain an image's subtle-ty of shadow, color, and motion.

With the advent of Pioneer's low-cost DVD-R drives and media, DVD-R is increasingly viable as a primary medium of distribution. For demo reels and corporate and industrial projects — even for independent features — craft-conscious folks from across the digital spectrum can finally (and thankfully) say goodbye to VHS or expensive film prints.

This month I look at DVD and MPEG-2 encoding from the shooter's or imagemaker's perspective.

With an eye on the DVD Studio Pro user, I examined several (relatively) inexpensive encoding and transcoding options from a range of manufacturers: Sonic Solutions, Apple, Media 100, Heuris, and Pixela. (For an in-depth review of DVD Studio Pro, click here.

Cleaner 5 SuperCharger ($999) is an elegant MPEG-2 solution for users already familiar with Cleaner’s easy workflow.

If you want to encode MPEG-2 from a tape source, you'll need a deck-controllable hardware encoder like the high-end Sonic Solutions SD-2000, which has a retail price of $24,499. You might also check out Media 100's MediaPress Pro, featuring realtime encoding and optional SDI, starting at $2,999.

Most DVD Studio Pro users will opt for a much simpler software solution, converting (transcoding) a QuickTime movie from Final Cut Pro or other NLE into the required elementary audio and MPEG-2 video files for authoring.

My goal was to explore the options available to serious imagemakers beyond the basic functionality of Apple's QuickTime-based transcoder.

As a frame of reference, I should tell you that my encoding standards can be quite rigorous. As a shooter for National Geographic Television since 1980, I have captured many images of wildlife, volcanoes, and water scenes. Their attendant high motion, subtle color variation, and deep shadows are the kinds of parameters that typically wreak havoc on MPEG-2 encoding.

To accommodate even the roughest video sources, SuperCharger offers four levels of noise reduction to remove snow and noise.

My concern for preserving the integrity of my images motivated me to explore the nuances of MPEG encoding. A few years ago, I became proficient on the Sonic DVD Creator system — a professional-class authoring package that allows a high degree of control over the encoding process. Using Creator, I have authored upwards of 40 discs in the last year, compressing and recompressing original footage from a range of sources from lowly VHS tape to widescreen 35mm film.

MPEG is a standard developed by the Moving Picture Experts Group (MPEG) to compress video and audio. MPEG video is similar to JPEG but has the extra ability to handle motion efficiently. While the intricacies of MPEG-2 are beyond the scope of this article, it is important for shooters to understand the general theory behind it.

The largest data element in MPEG is the Group of Pictures (GOP). This is a self-contained, decodable unit usually consisting of 13 or 15 frames. Dramatic compression gains are possible in MPEG because each GOP contains only a single complete frame, the remaining frames being mathematically derived from the initial I-frame (Intraframe).

These mathematically derived frames are both predictive and interpolative, meaning that MPEG has the ability to generate motion vectors that “see” ahead as well as fill in any presumed action that has been missed in the compression process, such as in the case of a bird flapping its wings.

Here is a typical GOP 13 pattern: IBBPBBPBBPBB.

SuperCharger’s adaptive filter suppresses the staircase aliasing effect known as “chiseling” and is particularly useful to DV shooters.

The I-frame is complete, and although compressed, it does not rely on another frame to be decoded. The P-frames are predictive as they look back at the previous I-frame for information. The B-frames are bi-directional, looking at both the preceding and following frames for data.

Quite simply, how well your encoder (or transcoder) accomplishes the predictive and bi-directional tasks will determine the quality of your encoding.

But encoders can and do get confused. Sharply defined, high-contrast lines or patterns may produce ugly halos or “mosquitoing” effects in footage. Fast-moving scenes tend to form large blocks at low bit-rates. With its high level of detail, DV especially will often display a hideous chiseled effect in MPEG-2.

Luckily, in the land of DVD, high bit-rates cure most, if not all, ills. The more bits assigned to describe a given block of pixels, the more accurately that block will be represented in compressed form.

Though speedy, Apple’s QuickTime transcoder does not offer sophisticated features such as inverse telecine or noise-reduction filters.

The highest video data rate allowed in DVD Video is 9.8Mbps. In Apple's iDVD entry-level product, the QuickTime transcoder is fixed at 8Mbps to ensure a high-quality result regardless of the complexity of the source material. (iDVD is bundled with many of the new Apple systems that include the DVD-R Super-Drive.)

The inefficiency of such a strategy becomes apparent when one considers that a single DVD-R 4.7GB disc can accommodate only an hour of content, more or less, at full bit-rate. To accommodate longer running times, a typical movie feature is encoded at 4.5Mbps to 5.5Mbps. Industrial or corporate projects may contain four hours of content or more, driving the average bit-rate much lower.

The question then becomes, what happens to the quality of our images as producers jam more and more content onto a 4.7GB layer?

Shooters must understand what's happening to their images at low and moderate bit-rates.

Many cinematographers use a small amount of smoke or haze, for example, to add texture and interest to otherwise static or boring scenes. But smoke and haze are precisely the kinds of elements that wreak havoc on MPEG-2 encoders. At rates less than 5Mbps, watch out. Welcome to Artifact City!

Sonic Solution's high-end hardware encoder, the SD-2000, features a plethora of options for interpreting problematic scenes. Users can alternate the block-encoding sequence or use a zig-zag pattern, apply advanced signal preprocessing, and seamlessly re-encode specific problem areas of scenes — like the wafting smoke.

Sometimes these tactics prove effective when nothing else will. In such cases, only the high-end encoder will do.

Image professionals looking for a more economical solution, however, will be looking at a software transcoder to accomplish their MPEG-2 goals. In the past, such transcoders have required inordinately long processing times — as much as an astounding 50:1 — often with less-than-satisfactory results.

Heuris MPEG Power Professional-2 (MPP2) offers a staggering array of high-end options, including the options of selecting advanced filters and setting maximum and target bit-rates.

Apple's QuickTime-based MPEG-2 transcoder has changed that considerably. Enabled in the Quick-Time Player as an export option after installing DVD Studio Pro, the deceptively simple-looking plug-in offers true, one-pass, variable bit-rate (VBR) encoding up to 9.8Mbps at hitherto unheard-of speeds.

On an older Apple G4/400, the MPEG-2 test reel with a running time of one minute, 15 seconds required five minutes and five seconds to transcode, or about 4X realtime. On the 733MHz G4 unsurprisingly, the required transcoding time was reduced considerably to three minutes and 20 seconds, or approximately 2.5X realtime.

But just how good is the QuickTime transcoder? For shooters looking only at QuickTime and DVD Studio Pro, this is a crucial question. Do we need a more sophisticated transcoder to achieve our image-quality objectives? Can we trust our toughest and most complex images to such a simple and inexpensive tool?

In most cases, the surprising answer is yes.

In my opinion, Apple's DVD Studio Pro is nothing short of revolutionary. Offering full-featured authoring to users for less than $1,000, DVD Studio Pro is attracting a wave of new adherents, offering shooters and Final Cut Pro junkies the ability to produce high-quality DVD Video titles on the same desktop used for everything else.

And when combined with the new Super-Drive DVD-R burner and low-cost 4.7GB general media, it's not surprising that DVD Studio Pro is transforming how DVD Video titles are being authored, and by whom.

Recently I put the Apple transcoder to the test, first to determine the nature of its limitations, and then to explore transcoder alternatives that could fill the gap.

This was fun. In Final Cut Pro, I assembled the toughest one-minute sequence I could muster — a compilation of potentially problematic shots from a mind-numbing new-age music video I shot for Tangerine Dream back in 1986.

Unique in its class, Heuris MPP2 ($1,499) features an analysis pass that injects I-frames at scene changes — a valuable feature typically found only in high-end encoders.

The sequence is an MPEG-2 encoder's worst nightmare. Originated on film at 24fps, the finished show contained as many as a dozen scenes from a Hawaiian rainforest layered one atop another. Flickering beams of sunlight through a bamboo canopy, cascading waterfalls, deep sun-dappled shadows, and sun-streaked rain falling through screens of flowers. It's enough to make even the best encoder flake out.

But Apple's new transcoder hung in there. The MPEG output looked quite good even at 5Mbps. Its most obvious artifacts — some haloing — could be attributed to the transcoder's lack of an inverse telecine feature.

Inverse telecine (IVT) removes the redundant fields introduced during the film-to-tape transfer. Unless these fields are removed, they will produce ugly artifacts such as haloing or mosquitoing, especially at lower encoding rates.

At 3Mbps, I discovered the practical limit of Apple's transcoder. In a test sequence sourced from 30fps DV, there was noticeable blockage in high-detail areas — in a series of window reflections, for example. And in one graphics segment, I noted a significant problem with chiseling: a staircase aliasing effect along the sharp edges of rapidly moving text. Some of this may be attributed to the nature of DV itself, but not all.

In my evaluations, Apple's QuickTime transcoder performed less than ideally at bit-rates of less than 5Mbps. For shooters, the lack of noise-reduction filters and inverse telecine is a significant drawback.

If you're already a Cleaner 5 user, Cleaner 5 SuperCharger addresses these shortcomings quite elegantly. As an economically priced hardware solution, I found its output markedly better than that of the Apple QuickTime transcoder at lower bit-rates (less than 5Mbps).

Strange but true, the Pixela Capty USB-to-MPEG encoder (about $300, currently only available in Japan) features the unique ability to edit MPEG-2 streams after encoding.

I should point out that SuperCharger, despite its hardware advantage, still requires almost twice as long as the QuickTime transcoder to perform the same task. The disparity seems to hold regardless of processor speed. In the world of MPEG-2 transcoding, there appears to be a time penalty for higher-quality results.

Seamlessly integrating the former Wired Digital Media Press into a familiar interface, the Cleaner 5 SuperCharger allows a full range of filtering, cropping, and processing options, in addition to inverse telecine and noise reduction. The noise reduction filters provide the ability to remove troublesome noise and snow in the source video. For shooters looking into DVD, this feature alone is enough to save a project and even a career.

Like SuperCharger, it may not be the fastest solution out there, but the Heuris MPEG Power Professional-2 might be just the ticket for shooters looking for truly exquisite results. If I were giving performance awards strictly on the basis of output quality, the Heuris MPP2 transcoder would win the prize hands down.

Still, compared to other solutions, Heuris MMP2 is inclined to take its time in producing first-class results. On a G4/400, it required more than 13 minutes to complete a file that the QuickTime transcoder could do in five minutes. Even compared to the SuperCharger, it took 25% longer.

But to shooters concerned first and foremost about the appearance of their images on DVD, the high-quality output of MMP2 may be worth it.

The latest version offers a full range of professional functions. It can perform an analysis pass to ensure optimal parameter settings. Source assets may be marked for start and end times, then concatenated into larger files. Key filters for suppressing noise and artifacts may be applied globally or to specific scenes — a cool feature for those of us who like to mix formats, say DV and BetaSP, in the same production.

Heuris MMP2 also features various signal processing filters, such as the FIR filter that reduces haloing around objects. A nonlinear filter (akin to the SuperCharger's adaptive filter) is useful to reduce the staircase chiseling effect — a nagging problem for DV shooters trying to mimic the look and feel of higher-end formats.

Heuris MPP2 offers DVD Studio Pro users a high-end transcoding option that is virtually indistinguishable from the most expensive hardware solutions. If you produce long-form DVD titles and can withstand the longer-than-average transcoding times, Heuris MMP2 is an excellent solution offered at a very reasonable price.

Toast 5.0 Titanium from Roxio allows DVD Studio Pro users to replicate discs from a DVD-R master.

A small Japanese company is marketing an ingenious MPEG-2 encoder for the Mac and PC. Although intended strictly for the consumer market, the Pixela Capty offers users the unique ability to encode analog sources in realtime via USB!

Understandably, the quality of encoding is less than optimal, as the software struggles to manage the onslaught of analog data via the relatively slow USB connection. Still, the Pixela Capty, which retails for $300, deserves closer scrutiny in at least one respect. It features an unusual “MPEG Cutter” that seamlessly removes unwanted GOPs from an MPEG stream after encoding.

One has to wonder why such a basic editing capability isn't available on high-end tools. In my own work, I could have used it many times to remove a distracting blip or unnecessary GOP between assembled assets.

At present, the Pixela Capty is only available in Japan. Check out the Pixela website (www.pixela.co.jp) for more information.

If you're an accomplished shooter and want to author your own DVD titles, chances are you're considering Apple's DVD Studio Pro. If so, you might want to look beyond Apple's capable-but-limited QuickTime-based transcoder.

You have a number of options. For ease of use and general workflow, Cleaner 5's SuperCharger offers unparalleled elegance in a familiar interface — plus a broad range of professional options, including inverse telecine and noise reduction.

Heuris MPP2 offers even more sophisticated filters and analysis features that consistently produce superb results.

With the advent of MPEG-2 camcorders like Hitachi's revolutionary DVD-CAM, encoders and transcoders may be unnecessary in the future, as original footage will already be DVD-compatible. In the meantime, for shooters turning to DVD Studio Pro, a knowledge of MPEG-2 and various transcoder options has become a necessity. MPEG-2 is now a fact of life, and if we are to survive as craft-savvy shooters in a DVD world, we must embrace it.

Barry Braverman is a veteran director of photography based in Los Angeles with more than 20 years of experience as a nonfiction specialist. He can be reached at bbrave@mindspring.com.

Shooters looking to encode their images to MPEG-2 will be looking primarily at variable-bit-rate (VBR) encoding. As opposed to constant bit-rate (CBR), which assigns the same number of bits to a scene regardless of content, VBR encoding affords scenes with more motion a higher bit-rate, while simple scenes with little motion get a lower bit-rate.

For related reviews of Panasonic's LF-D201U DVD-RAM drive and Apple's DVD Studio Pro, see reviews.