The Film Archive Forum - Understanding compression standards

These notes were produced by Luís Carrasqueiro, former Head of the Advisory Service for moving images and sound online at of the British Universities Film & Video Council (MAAS Media Online). They were prepared for the first Film Archive Forum training day, held at the Yorkshire Film Archive in September 2004. Documentation from future training days will be added to this site in due course.

Video (meaning moving picture and sound data for simplicity) is a data-rich medium. A raw video file is huge and impractical even for some local hard disks, let alone networks. The internet made image compression (or encoding) standards such as JPEG universal. Similar efforts were made in the moving image field and the first mainstream standard created for interactive applications delivered on CD-ROM was MPEG-1. Other compression standards followed (for a discussion of MPEG standards and nomenclature see the downloadable article in Further Reading below) and quickly multiplied, therefore diluting the meaning of the word 'standard'. MPEG-1, MPEG-2, MPEG-4, Sorenson, Motion JPEG, Cinepak, Indeo are all names of codecs. A codec (short for encoder/decoder) is a piece of software that is used both to encode moving pictures and sound and to then decode it and play it back to the final user.

It is difficult to say which codec is best since all have different applications. Even when a decision has been made on a codec, other factors need to be taken into account which depend on the intended use of the video file:

Frame rate – If the footage consists of a talking head reducing the frame rate may be acceptable and this will reduce the final file size. Fast action will require the frame rate of the original footage to be matched by the encoded version.
Frame size – How big should the encoded video look on screen? A crude approximation is that doubling the screen size increases the file size by a factor of four.
Bit rate – How much data is to be encoded per second? A 56 kilobit per second (Kbps) encoding will match a modem speed, thus, allowing for a quick download, but will look blotchy and with very low resolution. A 768 Kbps as is used by MAAS Media Online has a similar feel/resolution to VHS but creates much larger files. Digital TV is typically broadcast around 5 Mbps (megabits per second).

For those lost in the bit and byte jargon, please note that bit rates and connection speeds are given in bits (per second), whereas hard disk and memory sizes are normally given in bytes. For a guide on these units please see:

http://www.smartftp.com/support/kb/index.php/53
http://www.t1shopper.com/tools/calculate

A modem allows for a bandwidth of 56 kilobits per second (Kbps), a normal ADSL broadband connection 512 kilobits per second and an Ethernet network from 10 megabits per second (Mbps) to 1 gigabit per second (Gbps).

Codecs at work

What most codecs do is to divide each frame into blocks and look for degrees of variation within each block. For example, if a frame contains a large amount of cloudless sky which is just a flat blue area, the codec would not store information for every pixel but would instead just store a large square of a certain shade of blue. Storing information about a blue-filled square is far more economical than storing information for every pixel in the square.

Some codecs don’t encode every frame. For example, if the footage to be encoded is a long stationary shot of the 'Mona Lisa', the codec would not need to store 25 frames per second. Instead it would store one frame and replicate it for as long as the shot was stationary. This saves a very large amount of data.

The two processes described above are just two of the clever compression techniques codecs use. In summary, different application of video files require different compression methods (or codecs). A DVD needs high resolution and will use MPEG-2. Broadcasting video for low-bandwidth users would benefit from codecs such as MPEG-4, Sorenson, Windows Media 9 or Real 10.

Other factors that affect the physical size of the file have to be balanced: these are the frame rate, frame size and the bit rate.

Technology fragmentation – the user side

With so many different codecs, how can a provider guarantee that users will have the right technology (codec) on their computers to playback the video? There is no straightforward answer to this issue.

If a video file has been encoded using a Windows Media codec it is very likely that it will play on a Windows-based machine. However there is no guarantee that it will play on other platforms.

The main technologies are Windows Media Player – owned by Microsoft; QuickTime – owned by Apple Computers; and Real Player – owned by Real Networks. The latter is mostly used for streaming video, the first two are frequently used for download or streaming. For example, QuickTime can play files encoded with a variety of codecs but none in the Windows Media series. This is due to competition between the two technologies and licensing restrictions.

If a user’s system does not have a certain codec, it is possible that it will be available for purchase or even a free download. Player applications normally try to determine if one is available and will attempt to download the necessary codec. This is a significant problem as most educational institutions and large offices disallow installation of new software (which a codec is) without express permission and action by IT managers.

When choosing which codec to use for encoding video it is worth considering which platforms will be able to play it. MPEG standards are normally played by all platforms and applications but there is no 100% compatibility. MPEG-1 is the closest to universally playable but is not the most attractive option. This is because MPEG-1 cannot carry metadata (e.g. collection name, title, author). MPEG 4 is the new attempt at a metadata-friendly, 'universally playable' codec. MPEG was slow to nail down the specification and licensing conditions for MPEG-4, which means that MPEG-4 is now anything but universal.

MAAS Media Online has decided to encode in both Windows Media (Windows Media 7 codec) and QuickTime (Sorenson codec) formats, which is a more costly solution than producing a single format file (such as MPEG-4).

Protecting content

Archives often licence collections to third parties. If they are to do this digitally, steps should be taken to appease the rights holders and provide protection against unlicensed copying. Completely disallowing the copying of a video file is impossible, whatever the format, analogue or digital. What distributors or rights holders can do is to make such actions more difficult, troublesome and traceable.

Streaming versus download – see the downloadable article from Media Online Focus below.

DOG – a digital onscreen graphic, which MAAS Media Online uses, will visibly mark the provenance of the video file. Most television companies include their logos on screen when broadcasting their own productions. The DOG can be overlaid with another graphic, or the frame can be cropped to exclude the graphic. This is not strong protection but one that intensely discourages unlawful copying. Its advantage is that it is possible to trace provenance.

Watermarking – similar to using an invisible DOG. A watermark is embedded within the data of the video and is not necessarily visible. It is often possible to determine the presence of the watermark even if graphics are added to the frame. Interestingly, when an encoding is done, compression artefacts can be considered quite effective fingerprints. This isn’t extremely effective but can be used to help trace provenance.

Digital Rights Management – plenty of new developments in this area meant that it is becoming possible for the video objects themselves to be protected at individual copy level. As examples we have files which expire after a certain number of playbacks, or files which require a network connection that checks user authenticity and authorisation before playing the video. Ultimately this is every rights-holders dream but inconvenient for users.

The model used by the JISC and MAAS Media Online for Education Media OnLine (see Web sites page) content is one based on licensing and a degree of trust. User institutions sign a sub-licence which binds them to the terms of the 'mother-licence' with the rights-holder. The liability within the institution is then typically passed on to the user. Simple terms and conditions in the form of 'you may' and 'you may not' are shown and need to be accepted on every login. Authentication and authorisation is done via the ATHENS system, omnipresent in Higher and Further Education institutions in the UK. For more information please visit www.bufvc.ac.uk/maas/rights.

Further reading

These articles from the BUFVC's journal Media Online Focus are available as free downloads (PDF format):

Recordable DVD formats (776KB)

MPEG standards (885KB)

Editing using free software (895KB)

Streaming or Downloading (142KB)