6   Video Compression

Once you get past a few hundred kilobits-per-second, it's possible to deliver pretty good quality video and sound.

—Vinton Cerf

Video signals used in IPTV and Internet Video are almost always compressed. Compression means reducing the number of bits required to represent the video image. This is an important topic, because choosing a suitable compression method can sometimes mean the difference between the success and failure of a video networking project.

In this chapter, we will begin by examining the reasons for compression and look at some of the factors that determine what form of compression is suitable for an application. Then, we will examine MPEG video compression, since it is one of the most popular technologies used for video and audio compression. After that, we'll look at some of the other compression systems available for use with video and audio signals. We'll conclude with a look at some of the licenses needed to use some forms of compression technology.

Why Compress?

Many communication systems that have become commonplace in the past few years depend on compression technology. For example, MP3 players use compression to take files from audio CDs and make them small enough to fit into the memory of a portable player. Compression fits a two-hour movie onto a four-inch DVD. Both CATV and satellite television systems use compression to place multiple digital video channels into the space formerly occupied by a single analog video channel, allowing hundreds of video channels to be distributed economically to viewers.

Here are some of the main reasons why compression is used for IPTV and Internet Video systems:

•   Compressed streams can be transmitted over lower bit rate networks than uncompressed streams. For Internet Video applications in particular, this can mean the difference between getting the stream to a user or not. For example, many home users have Internet connections over standard ADSL lines that operate in the range of 1.5 Mbps. Unless a digital video stream is substantially compressed, it will not fit into this bandwidth.

•   More compressed streams can fit into a given bandwidth. This is particularly important for IPTV systems that have a fixed upper limit on bandwidth for a given distance. For example, ADLS2+ has a limit of just over 10 Mbps at a distance of 9,000 ft (2,750 m). With normal compression techniques, 10 Mbps has enough bandwidth for two or three video signals. As compression technology has advanced, more and more signals can be squeezed into the same amount of bandwidth.

•   Raw, uncompressed HD video signals occupy 1.5 Gbps of bandwidth, which is roughly 1,000 times the capacity of a standard ADSL link. Without compression, there would be no way to deliver HD video to a viewer over any of the normal IPTV networks.

•   A compressed video or audio file will occupy less space on a disk drive or other storage medium than an original uncompressed file. This enables users either to put more content in a given space or to use less space for a given file.

•   In many real-world video signals, there is a large amount of redundancy and underused bandwidth. Often a good portion of a single video frame is identical to the frame immediately before or after it. A good compression technique can use this redundancy to greatly reduce the amount of bandwidth.

Of course, there are compromises that must be made in order to achieve these benefits, such as:

•   Compression can introduce delay into a video or audio signal, at both the compression and decompression stages. This occurs because most video compression systems operate by looking at the differences between adjacent portions of the input signal, such as the changes from one frame to the next in a video signal.

•   Compression can be difficult on signals that have a lot of noise in them, such as static or other interference. When there is a lot of noise in a video signal, the compression system has difficulty in identifying redundant information between adjacent video frames.

Overall, the benefits certainly outweigh the drawbacks, particularly when you consider that IPTV and Internet Video providers really don't have a choice about using compression.

Groups of Pictures and Why They Matter

Users of any MPEG system will quickly encounter a variety of frame types, including I frames, P frames and B frames, as well as the term Group of Pictures (GOP). These terms all describe the way picture data is structured in an MPEG stream or file.

A frame is a single image from a video sequence. In NTSC, one frame occurs every 33 milliseconds; in PAL, one frame occurs every 40 milliseconds.

•   An I frame is a frame that is compressed solely based on the information contained in the frame; no reference is made to any of the other video frames before or after it.

•   A P frame is a frame that has been compressed using the data contained in the frame itself and data from the closest preceding I or P frame.

•   A B frame is a frame that has been compressed using data from the closest preceding I or P frame and the closest following I or P frame.

•   A GOP is a series of frames consisting of a single I frame and zero or more P and B frames. A GOP always begins with an I frame and ends with the last frame before the next I frame. The GOP is usually a fixed, repetitive pattern that is configured on the compression device. Different content suppliers may use different GOPs for different channels, but they are normally fixed within each channel.

To understand why MPEG uses these different frames, let's look at the amount of data required to represent each frame type. With a video image of normal complexity, a P frame will take two to three times less data than an I frame of the same image. A B frame will take even less data than a P frame—a further reduction by a factor of two to five. Figure 6.1 shows the relative amounts of data for each frame type in a typical MPEG GOP.

The Impacts of GOP Length

One parameter that system providers have a lot of control over is GOP length. Choosing the right length can be quite controversial.

Remember that a GOP always begins with an I frame. To determine the length of a GOP, you simply count the number of B and P frames between each consecutive I frame and add one for the I frame. For example, in the frame sequence shown in Figure 6.1, the GOP length is 12: one I frame, three P frames and eight B frames.

FIGURE 6.1   Relative Amounts of Data in Each MPEG Frame Type

A GOP is considered short when the GOP length is low, say three or five. Some systems use GOPs that are quite long; values of 15, 30 or even 60 have been used in some applications.

Selecting a suitable GOP length can have a big impact on a video network. Many system performance factors are affected by GOP size, including the bit rate of encoded streams, channel change time and the ability of the stream to tolerate errors. Let's examine each of these factors in more depth.

Bit Rate

As Figure 6.1 clearly shows, I frames contain more data than P frames or B frames. With a short GOP length, the total number of I frames in the stream is increased, thereby increasing the average amount of data that needs to be transmitted for each frame. This translates into heavier demand for bandwidth, which can affect the performance of both IPTV and Internet Video services. With longer GOPs, there are fewer I frames per second, so the aggregate data rate drops.

Channel Change Timing

Whenever channel changing occurs in a video stream, the decoder has to have enough data to accurately produce a new image sequence. The decoder's ability to do so depends on which type of frame it receives. If the decoder receives an I frame, then no problem, because each I frame contains all of the data to completely reproduce one frame of video. If the decoder receives a P frame or a B frame, then it has a problem, because these frames only contain enough data to tell the decoder about any changes that have happened since an earlier frame. So, what typically happens after each channel change is that the decoder waits for the first I frame of the new video channel to arrive before it begins to produce an image.

With a short GOP, of, say, five frames, channel changing isn't much of a problem. In a 30-frame-per-second system (such as those used in the U.S.), this means that the decoder needs to wait, at most, 166 milliseconds for the first I frame, and that amount of delay is insignificant to viewers. If, on the other hand, the GOP is 30 or 60 frames long, it could mean that the decoder may need to wait one or two seconds before the first I frame arrives. This can be quite annoying to viewers.

Two different approaches have been demonstrated to address this issue. One method uses a server that stores decoded copies of all the videos present on an IPTV network. When a user changes channels, the STB momentarily connects to the server to get a sequence of I frames for the new channel and then rejoins the regular long GOP stream once a new I frame is delivered. This approach can deliver very fast change times, but there are some questions about how well this technology will scale to thousands of users all changing channels simultaneously during a major live sporting event.

Another system actually makes available two versions of each stream for use by STBs—one with low resolution and a short GOP, and another with normal resolution and a long GOP. Normal viewing is with the long GOP, normal resolution stream. When a channel change occurs, the STB connects to the low-resolution stream and upconverts it to a normal size picture. Once the normal stream is ready (i.e., when an I frame arrives), the STB switches back to the normal stream.

This method has the advantage of not requiring any special servers or targeted streams to be delivered to each STB, but it does require two versions of each stream to be available. The low-resolution streams can also be used for picture-in-picture applications when they aren't being used for channel changing.

Error Tolerance

One major benefit of an I frame is that it permits the STB to wipe out any memory that it has about previous frames. This contrasts with P and B frames, which require the STB to store a copy of the preceding frames so it can properly create the new frame. Consider what happens if one of the incoming frames in the middle of a GOP has an error. This error can persist in the STB for a while, until the next I frame arrives. Once this happens, the error can be cleaned out.

MPEG

The Moving Pictures Experts Group has developed some of the most common compression systems for video around the world and given these standards the common name of MPEG. Not only did this group develop video compression standards including MPEG–1, MPEG–2 and MPEG–4, but it also developed audio compression standards, which we will discuss later in this chapter. This group continues to meet and to set new standards (which we won't discuss here), including work on MPEG–7 (a standardized means for describing audio and visual content) and MPEG–21 (standards for describing content ownership and rights management).

MPEG standards have enabled a number of advanced video services. For example, MPEG–based DVDs have replaced the videotape as the preferred medium for viewing Hollywood movies in the home. Digital television, including digital satellite television and digital cable television, is based on the MPEG video compression standards. High definition (HD) television also uses MPEG technology. Also, much of the content for streaming media on the Internet is compressed using MPEG or closely related technologies.

MPEG–1

MPEG–1 was the first standard developed for video compression by the Moving Pictures Experts Group. It was intended for use in creating video CDs, which had some popularity in computer multimedia, but never completely caught on as consumer movie rental or purchase format. MPEG–1 is still in use today for a number of applications, including low-cost surveillance cameras and some Web video applications. It is also interesting to note that MPEG–1 is allowed as a video compression method for DVDs, and many DVD players will play video CDs. Stand-alone and software-based MPEG–1 encoders are available for very reasonable prices from several sources. MPEG–1 does not support interlacing, so standard full-resolution PAL and NTSC signals are not usable with MPEG–1.

MPEG–2

MPEG–2 is the predominant standard for MPEG video today. It is used in a wide variety of applications, including digital TV production and broadcasting, HDTV, satellite television and cable television. Each day, thousands of hours of MPEG–2 video are recorded, processed and played back by television broadcasters around the world. Millions of hours of MPEG–2 recordings are sold to the general public each day in the form of DVDs. Thousands of PCs with MPEG–2 playback capability are sold each week, and the installed base of MPEG–2 devices continues to grow.

MPEG–2 supports standard NTSC and PAL signals at full resolution, as well as 720p and 1080i HD signals. MPEG–2 also enables multiplexing of a number of video and audio streams, so applications like multi-channel satellite television become possible. MPEG–2 also supports five channel audio (surround sound) and the Advanced Audio Coding (AAC) standard.

Many MPEG–2 devices, including highly sophisticated MPEG–2 encoder and decoder devices are available as custom semiconductors; these are in their third or fourth generations. There are literally hundreds of millions of STBs, digital satellite receivers and DVD players installed in consumers’ homes that can decode MPEG–2 signals. A wide variety of MPEG–2 equipment is available for functions such as statistical multiplexing, bit rate converters, telecom and IP network adapters, and more.

Various software tools are available for producing MPEG–2 streams using general-purpose PCs. With sufficient processing power and memory, a PC can be used to create an MPEG–2 stream in real time. However, for many applications, such as program editing and production, real-time performance is not necessary, and even moderate-performance PCs can create MPEG–2 compressed video files for later playback.

Software-only decoders for MPEG–2 exist, but they can be difficult to run without the addition of a hardware accelerator to a desktop PC. Adding an accelerator can drive up the cost and complicate the deployment of networks intended to deliver video streams to a large number of desktops or homes.

Overall, MPEG–2 is a well-defined, stable compression system with a wide variety of applications. Hundreds of millions of devices installed around the world are capable of receiving and decoding MPEG–2 video in a wide variety of flavors. MPEG–2 is commonly used in contribution, distribution and delivery networks. However, MPEG–2’s video and audio quality are not competitive at stream rates below 2.5 Mbps.

MPEG–4

MPEG–4 is a much more recent product of the standards process, the first version having become formally approved in 2000. As would be expected, MPEG–4 incorporates a whole range of new technologies for video compression that have been developed in the past decade. All of the necessary support technology for MPEG–4 systems, such as custom semiconductors, is being developed to widely deploy MPEG–4. MPEG–4 AVC may also make it possible for high definition signals to be encoded at bit rates below 10 Mbps, opening up a much bigger range of technologies for transporting HD video signals.

Prior to the introduction of the MPEG–4 Advanced Video Coding (AVC) standard, MPEG–4 did not offer truly dramatic performance improvements over MPEG–2 for compressing live natural video sequences, including most types of news, entertainment and sports broadcasts. Basic MPEG–4 has a number of advantages for synthetic (computer-generated) video and has already deeply penetrated IP video streaming applications (e.g. Apple's QuickTime has fully migrated to MPEG–4). Most desktop PCs can already decode MPEG–4 video using media player software that is freely available on the Internet.

MPEG–4 AVC is a more recent offering (circa 2004) and has the potential to replace MPEG–2 in the long run. The reason is that MPEG–4 AVC can achieve quality levels that compare favorably to MPEG–2 at half the bit rate. Of course, there is a cost to this, in terms of the greater processing power needed to encode and decode AVC signals. In addition, because AVC is newer, the technology has not had a chance to pass through as many learning and optimization cycles as MPEG–2 has undergone since 1996.

One potential drawback of MPEG–4 is that decoders are more complex for MPEG–4 than for MPEG–2. According to the MPEG–4 Industry Forum (www.m4if.org), an MPEG–4 decoder will be 2.5 to 4 times as complex as an MPEG–2 decoder for similar applications. This means more complicated (and therefore more expensive) hardware devices and greater demand on processor resources for software decoders. Before the decision is made to use MPEG–4 in a video delivery system, it is important to test any user devices (STBs, desktop PCs, laptops, etc.) that will be used to decode the video signal. Service providers may need to avoid using some of the advanced features of MPEG–4 and stick to the simpler profiles. Also, service providers may need to work closely with STB vendors to ensure that adequate supplies of decoder chips are available to meet deployment schedules.

Overall, MPEG–4 is an exciting new collection of technologies that promises to greatly increase the amount of video information that can be squeezed into a given amount of network bandwidth. Through MPEG–4 AVC, much more efficient video coding is possible, and the variety of object types available makes integration with computer-generated graphics simple and extremely bandwidth efficient. Because of MPEG–4’s complexity and its relative newness, much development work needs to be done to reach the level of sophistication and maturity enjoyed by MPEG–2 technologies.

Audio Compression

Just like video compression, MPEG has a variety of audio compression options. There are three layers of MPEG audio (conveniently called Layers I, II and III) and a newer audio compression standard called Advanced Audio Coding (AAC). In this section, we'll take a short look at each one of these. Note that any of these audio compression methods will work with any type of MPEG video compression, except that MPEG–1 streams do not handle AAC audio.

MPEG audio Layer I is the simplest compression system. It uses 384 input samples for each compression run, which corresponds to 8 milliseconds of audio material using 48 kHz sampling. Each band is processed separately, and then the results are combined to form a single, constant bit rate output. Layer I can achieve a compression ratio of 4:1, which means that a 1.4 Mbps CD-quality audio signal can be compressed to fit into a 384 kbps stream with no noticeable loss of quality. Compression beyond this—to 192 or 128 kbps—results in lower quality.

MPEG audio Layer II uses more samples for each compression run, 1,152 to be exact. This corresponds to 24 milliseconds of audio at 48 kHz sampling. This enables frequencies to be resolved more accurately. Layer II also eliminates some of the redundancy in Layer I coding, thereby achieving better compression, up to 8:1. This means that CD-quality audio can be achieved with a stream rate of 192 kbps.

MPEG audio Layer III uses the same number of samples as Layer II, but it uses them more efficiently. Layer III has an audio mode called joint stereo, which capitalizes on the strong similarities between the signals that make up the left and right channels of a stereo program. It also uses variable-length coding to more efficiently pack the compressed audio coefficients into the output stream. As a result, Layer III encoders can pack CD-quality audio into streams as small as 128 kbps, achieving compression ratios as high as 12:1.

MPEG AAC is available only with MPEG–2 or MPEG–4 video streams. It supports up to 48 audio channels including 5.1 audio. Very good quality results for surround-sound applications can be achieved with AAC at 192 kbps.

Dolby AC-3 Audio

Dolby AC-3 audio coding is also commonly known as Dolby Digital. It offers a high-quality audio experience with good compression characteristics and has been approved for use in both DVDs and in digital television broadcasts in the U.S. Dolby AC-3 audio is included in some versions of MPEG–4 and is used on a number of satellite television systems.

Overall, MPEG audio is flexible and does not require near the magnitude of processor involvement of MPEG video. As the layer number goes up, the complexity of both the encoder and the decoder go up, but so does the compression ratio. Software-only Layer III decoders can run smoothly in a wide variety of personal computers. AAC decoders are not as common, possibly due to the complexity involved. When choosing an audio-encoding method, remember that the overall transport bandwidth must be high enough to carry the video signal, the audio signal and some overhead to make the streams operate correctly.

Microsoft Windows Media and VC-1

Windows Media Player has a long development history from Microsoft. With a recent version, Microsoft took two unusual steps. First, Microsoft pledged to freeze the video decoder design for a number of years, to provide an incentive for semiconductor and other hardware device manufacturers to spend the time and resources necessary to incorporate Windows Media into a variety of low-cost products. Second, Microsoft won approval of the video decoder design as a public standard named SMPTE 421M (informally known as VC-1) from the Society of Motion Picture and Television Engineers. Now, any company that wishes to design a VC-1 decoder will be able to do so, provided that it obtains a license to use any of the patented intellectual property in the specification that belongs to Microsoft or other parties.

Microsoft intends to address a broad cross-section of the video compression market with VC-1 technology. The company has already released a number of implementations of its technology that range from low bit rate streaming for hand-held devices all the way up to digital projection of first-run theatrical motion pictures. In addition to the VC-1 video-encoding technology, Windows Media covers other aspects of the complete package, including audio coding, stream formats and DRM. In addition, the company is very aggressively pricing licenses to make VC-1 attractively priced relative to other technologies, such as MPEG–4.

Some readers may wonder about the differences between VC-1 and MPEG–4 AVC. Both codecs offer significant advances in coding efficiency (i.e., fewer bits for a given picture quality) as compared to MPEG–2. To date, there hasn't been any compelling evidence to say that one is clearly better than the other for any large group of applications. Interestingly, many vendors of encoders and decoders are designing their hardware to support both technologies, through the use of general-purpose digital signal processing (DSP) hardware and downloadable firmware.

Other Compression Technologies

MPEG and Microsoft are not the only games in town. Here are a few other compression technologies that bear consideration for service providers, primarily those in the Internet Video market.

JPEG

Standards developed for compressing still images by the Joint Photographic Experts Group are named JPEG files. These standards have been adapted for video use by treating each frame of video as a separate picture and compressing it. The approach brings some benefits, most importantly is the ease in which motion sequences can be edited. Since each frame of video is compressed individually, there are no structures like the GOPs of MPEG and therefore no restrictions on when one frame sequence can be stopped and another started. JPEG files are used in some video editing systems precisely for this reason.

JPEG2000

JPEG2000 is an advanced form of still image compression that was finalized in 2000 (hence the name). It uses a completely different technology for image compression than JPEG, but performs the same tasks. JPEG2000 also compresses each frame of video individually, so the technology is not able to take advantage of the similarities between adjacent frames. As a result, streams tend to be higher bandwidth than those commonly used in IPTV and Internet Video applications.

Proprietary Codecs

A number of proprietary video and audio codec systems are on the market, and many of them are suitable for use in Internet Video networks. Because they are proprietary, the exact details of their operation are normally not provided for general publication. In addition, the different codec manufacturers are currently engaged in heated competition, so product cycles are short and performance and other specifications can change rapidly. Let's look at two of the largest codec suppliers for the video streaming market: Real Networks and Apple.

Real Networks is a major supplier of proprietary codec technology. Most of Real's products are targeted at the video streaming market, but more developments are sure to come. As with Microsoft's products, a number of third-party tools (from suppliers such as Adobe) can be used to create compressed video streams in both real-time and off-line production environments. A good deal of content is available for streaming on the Web in Real's SureStream format, which is designed to automatically adapt to suit the wide range of different network connection speeds used around the globe.

Apple Computer supplies QuickTime technology, which has migrated to using standards-compliant technology such as MPEG–4. Apple was one of the pioneers of video streaming and still has a significant amount of development activity underway for new technology.

One distinguishing feature of both of these codec suppliers is their willingness to provide a free software client (player) for receiving their compressed video streams. Literally hundreds of millions of personal computer users have downloaded and installed these players onto their desktop and laptop computers. In addition, most of these companies also supply a free encoder with limited functionality. More sophisticated encoders are generally available for a fee; these versions often contain advanced features that can make the job of creating content files easier, as well as using more efficient compression algorithms.

There are no easy answers when deciding whether or not to use proprietary codecs. All of the main software-based codec suppliers mentioned in this section have a long and distinguished track record of innovation and customer service. The same can be said for many hardware-based codec suppliers. Nevertheless, any users of a proprietary codec run the risk that their supplier will, for one reason or another, stop providing products. Prudent users will assess this risk and have a contingency plan in place. Here are some advantages and disadvantages of proprietary codecs:

Advantages

•   Innovation: As compression technology advances, innovations can be incorporated into proprietary codecs very rapidly. Industry standards tend to have a slower rate of change because of the need to achieve agreement between many different parties.

•   Pricing: Many proprietary software codec suppliers offer basic versions of their players (decoders) free and have very low cost encoder options.

•   Backward Compatibility: Proprietary codec suppliers have a strong incentive to ensure that new versions of their codecs work with previous versions and have typically done a good job in this area. This may not be as true with designs based on standards, unless backward compatibility is explicitly defined in the specification.

Disadvantages

•   Portability: Because a single vendor controls when and how proprietary codecs are implemented, versions for alternative platforms may be late to arrive or never produced. This can limit users’ choices, particularly in the selection of operating systems.

•   Change Control: Major codec suppliers determine when new features are released to the market and frequently encourage end users to upgrade to the latest version. This can make it difficult for large organizations to ensure that all users have the same version and to ensure that the codec software doesn't interfere with other applications.

•   Platform Requirements: As codecs become more powerful, the minimum requirements for other system components (operating systems, processor speeds, etc.) can also increase. This can force users to deploy system upgrades in order to use the latest versions of some software codecs.

•   Archival Storage: As with any rapidly evolving technology, long-term storage of encoded video files is useful only as long as suitable decoder software is available. In the case of proprietary codecs, the supplier controls software availability over the long term.

Digital Turnaround

Digital turnaround is the process of taking video and audio signals that are encoded in one format and converting them into another format. This normally occurs under the control of a service provider to help standardize the operation of a multi-channel system. If each stream has the same compression technology, GOP length and bit rate, then the process of channel changing is greatly simplified. One fixed-bandwidth stream simply replaces another stream of the same bandwidth whenever a viewer decides to switch. Digital turnaround usually consists of two tasks: transcoding and transrating.

Transcoding is the process of converting a video signal that is encoded in one technology (say MPEG–2) into another technology (say MPEG–4). The best quality results can usually be obtained if the signal is never fully decompressed or recom-pressed, enabling the output signal to retain some of the information embedded in the original video feed.

Transrating is the process of changing the bit rate of video streams. Most IPTV providers convert all of the incoming content into a common bit rate, using one rate for all SD content and a second rate for HD content. Transrating needs to happen frequently, because most content suppliers use a higher bit rate for distributing their content than the rates that most IPTV and Internet Video service providers choose to use.

Reality Check

In this chapter's Reality Check, we discuss the licenses necessary to use some of the compression technologies we have described here. Many readers may not be aware of this, but every DVD player and every DVD disc sold includes the cost of a mandatory license fee collected for each unit produced. Service providers need to consider license terms when analyzing the costs of installing a video delivery system.

Technology Licensing

As we have seen in this chapter, a huge number of clever technologies have been applied to the art and science of video compression. Even though much of this technology is governed by international standards, not all of this technology is in the public domain. In fact, many of the key technologies used in MPEG and other compression systems were developed by individuals and corporations who still retain ownership of their technology in the form of patents and other legally protected rights. For example, the patent portfolio for MPEG–2 technologies includes 630 patents from around the world.

Fortunately, the owners of these technologies banded together to set up an organization known as the MPEG LA (the LA originally stood for Licensing Administrator, but now LA is the official name). MPEG LA is responsible for establishing and collecting the license fees on the technology and for distributing the collected funds to the patent owners. This central clearinghouse provides big benefits to the users of this technology, because one simple payment to MPEG LA satisfies the patent obligations for the covered technology. Contrast this with the headaches and complexities that would be involved in negotiating separate license agreements with the 20+ companies that have patents included in the MPEG–2 technology pool.

The license fees are assessed on a per-item basis and are officially described on www.mpegla.com. For example, the fee listed on the Web site for an MPEG–2 decoding device (such as a DVD player, STB or computer with a DVD player, whether hardware or software) produced after 2002 is U.S. $2.50. Other fees are assessed for MPEG–2 encoders, MPEG multiplexers and other devices. Fees are also assessed for recorded media, such as DVDs, but the fees are relatively low (e.g., $0.03 for a single-layer DVD disc—although there are a number of different ways of calculating the fee).

There are similar fee arrangements for MPEG–4 devices. In addition, there are fees based on the number of streams created and on the number of subscribers served in cable and satellite television systems. In addition, there are fees for individual titles sold to viewers on a DVD or via pay-per-view, such as a VOD system. These fees have created some controversy in the industry, because they include charges for the device itself (like MPEG–2) and also charges for viewing content using the device.

Where does this leave the owner of a video networking system? First, it is important to understand that fees for devices are normally collected from the device manufacturers, so end users of equipment generally don't need to worry about technology fees. Second, publishers of media, such as DVDs, are also responsible for paying the fees required for those items. Third, most of the MPEG–4 license fees that are payable on a per-stream or a per-subscriber basis are targeted at companies that are charging users to view the videos. However, this arrangement will be subject to revision in 2008, so users of MPEG–4 would be well served by investigating the necessary license arrangements in detail before launching a large-scale system.

Summary

Video compression is a requirement for essentially all IPTV and Internet Video systems. We began this chapter with a discussion of why compression is so important. GOP length, a very important topic for service providers to consider, was discussed in depth. We then took a look at the varieties of MPEG for both video and audio applications, as well as other compression systems, including Microsoft's VC-1, JPEG, and offerings from Real Networks and Apple. We concluded with a brief look at digital turnaround and a discussion of licensing issues.

Any service provider needs to make a careful evaluation before choosing a compression technology. Each technology has benefits and drawbacks in terms of performance, cost, availability and scalability that can have major impacts on business plans, deployment schedules and viewer experiences. This choice cannot be taken lightly, because providers will need to live with their choices for years to come.

 

1. Interactive TV Today blog, August 31, 2005, blog.itvt.com/my_weblog/2005/08/eric_ cooney_pre.html