• Blogs
• Forums
• Tradeshows
• Partners
• White Papers
• Events

Optimizing Video for Streaming

Sep 26, 2001 12:00 PM, By Philip De Lancie

To improve the quality of streaming video, you need to know what options are available to make things better, and when to use them.

If you work with moving pictures — video or film — you probably spend a good chunk of your waking hours thinking about image quality: how to achieve it and how to maintain it. So it may not be easy to get excited about a medium in which the label “near-VHS-quality” actually has a positive connotation.

But like it or not, the streaming of video over networks is here to stay, another piece of the ever-expanding media pie that keeps us all fed. As broadband deployment grows to cover (eventually) the majority of American homes and businesses, more and more viewers will be watching their video streamed. And no one has a bigger stake than the production community in ensuring that the quality of those streams is as high as it can possibly be.

Obviously, the key factor driving streaming video’s popularity is convenience. “It’s ‘what I want, when I want, where I want,’” says Randy Thomas, director of marketing at Teranex, Inc. in Orlando, Florida, which develops compression algorithms and tools for maintaining the best possible image-quality within available bandwidth. “It all comes down to video-on-demand.”

Although many have claimed that online movies is the ultimate killer app for streaming, the reality at the moment is that the greatest -- and most profitable – use of streaming is coming from the area of education and corporate communication. With streaming, every school, institution, and corporation can have, in effect, their own personal broadcasting network.

“Education,” says Matt Cupal, senior VP for strategy and business development at Sorenson Media in Salt Lake City, Utah, the company that developed the video codec used in QuickTime streaming, “will benefit so much more from readily available video because much of what we learn is from personal interaction with and observation of a teacher. You can't see the countenance of a HTML page.”

Likewise, such benefits can be realized in the workplace. “In the corporate market, where there is plentiful bandwidth on intranets and virtual private networks, streaming is being used increasingly to communicate with company employees,” says Greg Lowitz, general manager of the webcasting solutions group at Pinnacle Systems in Mountain View, California, whose broadband production and Web encoding solutions include StreamGenie and StreamFactory. “Whether it's the CEO talking about recent financial performance, or a product manager launching a new software release, streaming helps people communicate their message quickly, convincingly, and with impact, particularly in corporations with remote offices around the world.”

Maintaining Standards

Once we accept the utility of streamed video in various fields, it’s realistic to recognize that untrained viewers probably won’t be in the front lines demanding higher quality. Instead, technical and creative professionals will have to advocate for the maintenance of quality standards through to delivery, especially if bottom-line business types seem ready to settle for less. Of course, to be an effective advocate, you need to know what options are available to make things better, and when to use them.

It would be convenient if there were a single step at which vigilance is important, or a single process that magically makes all the difference. But like any other production process, the preparation of video for streaming is a chain that is only as strong as its weakest link. “Each part of the chain is crucial to the success of any streaming event,” Lowitz observes.

“If you don't have the expertise to encode well,” Cupal adds, “you'll kill the content before it even gets off your machine. Delivery is also key. The Internet is a cold, dark place where many a packet can get lost. A great delivery network is key to minimizing this danger. And of course viewer-bandwidth has a tremendous impact on final viewing quality. 300kbps means near-VHS quality; 28kbps means postage-stamp. Finally, preparation of the video is important. It’s key to know how to shoot content for online consumption as well as how to properly store it.”

Lowitz says the guiding principle in optimizing video for streaming is to maintain "contribution quality" video and audio all the way up to the point of encoding. “Where possible,” he says, “use component video, either S-VIDEO, DV, or SDI. This provides the cleanest video image without artifacts due to dot crawl. For audio, use balanced connections to keep noise and hum to a minimum. Make sure levels are optimized for the best signal-to-noise ratio. And use a compressor/limiter, if available, to keep audio levels consistent.”

If you are actually shooting your content with streaming in mind, rather than simply preparing existing footage, you’ve got even more variables to work with. Good practices begin with understanding how the video compression algorithms (codecs) do their job.

“Compression basically works by finding high correlation between pixels that are adjacent in either space or in time,” Cupal says. “More similarity means more correlation means better compression.”

Asked for basic shooting tips, Cupal says to “use a good-quality camera, use a tripod, shoot lower motion shots, and use solid colors rather than patterns in clothing and sets. Avoid slow pans, and avoid shooting outside with trees as background.”

Pre-processing

Once the video is captured, there is still much to do to make it work better for streaming. “Depending on the target bit-rate, pre-processing of the audio and video image prior to encoding can significantly improve video quality,” Lowitz says. “While the latest codecs from Real and Microsoft are already excellent, additional pre-processing can improve quality another 20 to 30 percent.”

“You need to takes steps to ensure that you are not encoding information that is not required in the final output,” says Curtis Palmer, CTO and co-founder of Sonic Foundry, the Madison, Wisconsin company whose media creation and editing tools include Vegas Video, Sound Forge, and ACID. “You need to start by capturing the best-quality source video possible from a properly-calibrated system at full resolution.”

If the content originated on film, Palmer continues, inverse telecine should be used to bring the video back to its true frame rate. He also suggests running an intelligent de-interlace algorithm, if needed. “The idea is to ensure you have a progressive scan source before encoding it in any streaming media format.”

“Inverse telecine is quite important for film content at any bit-rate,” Thomas concurs. “By removing the redundant fields inserted during 24-to-30 fps film-to-NTSC conversion, a guaranteed 20-percent bit-rate reduction may be achieved with no loss of picture quality. If encoding the 24fps signal requires 20-percent less bits, the bit-rate can be turned up by 20-percent, improving the picture quality accordingly. The trick is to remove those redundant fields in the midst of bad edits, broken cadence, and orphan fields. It takes quite a bit of processing power to address all of the possible permutations of mixed 3:2 pull-down.”

The next step would typically be noise reduction. “Noise reduction helps at all bit-rates,” Thomas says, “but more so at lower bit-rates where every bit counts. By removing information that is not important relative to the content, a significant number of bits may be saved in the stream. Saving bits results in less of the visible picture content being discarded during the quantization process.”

Once you have the cleanest source possible, Palmer advises cropping the source video to remove areas that do not need to be encoded. As examples, he points to both letterboxing and the overscan area that would not normally be visible on a standard television. “Streaming media can display 16:9 aspect ratios without letterboxing,” he says, “so why feed this to the encoder?”

The final steps he suggests are to “use a high-quality scaling algorithm to reduce the frame size and adjust for non-square pixels from the video source. Also, a high-quality frame rate converter may be required for low bit-rate encodes. Note that with the proper tools, such as Vegas Video, many of these steps can be accomplished in a single pass.”

For audio, Palmer says, the steps are analogous to video: “Capture the best source you can. Clean it up if needed. Adjust the levels and use a peak limiter if needed. Use a high-quality sample rate converter if you need to reduce the sample rate.”

Choosing the Tools

Tools to optimize video before compression are available in many forms. They are frequently built into encoders, including software such as Media 100’s Media Cleaner Pro, the optional HiQ module for Pinnacle’s StreamFactory, or standalone hardware such as Teranex’s StarStream Compression Station.

One factor influencing the appropriate type of tool to use is whether the streaming will be on-demand or real-time. “With live content,” Cupal says, “you’re constrained to the connection’s bandwidth at any given second, as well as the ability of the video encoder’s CPU to keep up. Today’s CPUs max out at roughly 320 x 240 at 30fps in real-time. This works well for most applications and can be VHS quality, but it's not the quality people are getting from our high-end VOD stuff. With VOD you can play tricks like two-pass compression that do wonderful things for your video quality.”

“For VOD, human-assisted multi-pass encoding enables the balancing of bit-rate versus subjective quality,” Fritsch agrees. “Live Webcasting is more difficult. You don't have the luxury of sophisticated pre-processing, and you don’t get a second chance to encode the content.”

Live or on-demand, much of the optimization process is dictated by the target bit-rate. “Know your audience,” Thomas says, “and the distribution bandwidth to them. This will determine how much and what kind of optimizing you will need to do.”

As for specific features, Thomas says the most important to look for are “intelligent inverse telecine, high-order down-sampling filters rather than just field deletion, and de-interlacing that takes motion into account. Automatic scene-change detection can also help by placing the key frames in the optimal location for coding efficiency.”

Of course, the equipment by itself can’t do everything; it also takes the judgment of an experienced operator. “Even though tools have improved considerably in recent years,” Palmer says, “they still require that the user know how to use them to get the best results. A little knowledge about preparing content for encoding to streaming media can go a long way. On the other hand, not knowing how or when to use certain tools can lead to disastrous results.”

Lowitz agrees. “With so much focus on the technology behind streaming,” he says, “it's easy to forget that a production is only as good as the talent, the production values, and the content. In particular, since Web video still has certain limitations in the delivery of very high-quality images, even more attention is required on the production itself to optimize the limited bandwidth available, and to create a video stream that is both compelling and engaging for the viewer. So the talent behind the scenes can rest assured that their skills are now more important than ever.”

Phil DeLancie is a freelance writer based in Berkeley, CA.

© 2008 Penton Media, Inc.