Bitrate peeling

From Wikipedia, the free encyclopedia

Bitrate Peeling is a technique used in Ogg Vorbis audio encoded streams, wherein a stream can be encoded at one bitrate but can be served at that or any lower bitrate. Bitrate Peeling is similar in overall effect to Real Networks' technology called "SureStream", which basically creates a single media clip that contains multiple bitrates. The purpose is provide access to the clip for people with slower Internet connections, and yet still allow people with faster connection to enjoy the higher quality content. The server automatically chooses which stream to deliver to the user, depending on user's connection speed.

As of 2005, Ogg Vorbis bitrate peeling exists only as a concept as there is not yet an encoder capable of producing peelable datastreams [1].

An experimental implementation of bitrate peeling can be found here. However, re-encoding files at a lower bitrate will preserve more quality than this experimental bitrate peeler.

1 Difference from other technologies
- 1.1 SureStream example
- 1.2 Bitrate Peeling example
2 See also
3 External links

[edit] Difference from other technologies

The difference between SureStream and Bitrate Peeling is that SureStream is limited to only a handful of pre-defined bitrates, with significant difference between them, and SureStream encoded files are big because they contain all of the bitrates used, while Bitrate Peeling uses much smaller steps to change the available bitrate and quality, and only the highest bitrate is used to encode the file/stream, which results in smaller files on servers.

[edit] SureStream example

A SureStream encoded file is encoded at bitrates of 16 kbit/s, 32 kbit/s and 96 kbit/s. The file will be about the same in size as three separate files encoded at those bitrates and put together, or one file encoded at the sum of those bitrates, which is about 144 kbit/s (16+32+96). When a dial-up user has only about 28 kbit/s of bandwidth available, the Real server will serve the 16 kbit/s stream. If the dial-up connection is of higher quality, and maybe about 42 kbit/s is available, the server will automatically switch to the 32 kbit/s stream. A DSL or Cable Internet user will be served the 96 kbit/s stream. This looks good, but even though the user with 28 kbit/s can use a higher bitrate / higher quality stream (maybe 22 to 24 kbit/s), such thing can't be done with SureStream, unless the encoded file contains such a bitrate. This is where Bitrate Peeling comes into play.

[edit] Bitrate Peeling example

Contrary to SureStream, Bitrate Peeling requires only the highest bitrate to be used when encoding a file/stream, which is 96 kbit/s in this case. The obvious benefit is much smaller space on a server required by such a file. An additional feature of Bitrate Peeling is a much finer tuning of available bitrate/quality. Now, if a dial-up user with 28 kbit/s available bandwidth connects to an Ogg Vorbis file/stream, the server will "peel" the original 96 kbit/s file/stream down to the available bandwidth, the full 28 kbit/s (or 20 to 24 kbit/s, as overhead reduces the effective bandwidth). This "peeling" process is different from transcoding because transcoding uncompresses the file and recompresses it (a computing-intensive process), whereas the peeling process "cuts the fat" by removing excess bits from the stream without using much processor time. If 42 kbit/s is available, the server will peel the 96 kbit/s down to the maximum available bandwidth, maybe about 36 to 40 kbit/s. The benefit of Bitrate Peeling, besides space savings / smaller files, is smaller steps in bitrate and quality, which means that whatever the currently available bandwidth, the streamed file will be of highest bitrate and quality possible for that bandwidth. These benefits are only theoretical, as the only Vorbis peeler available is still in experimental state and produces file qualities inferior to what transcoding the higher bitrate file to a lower bitrate would.