Previous: Hardware decode/mux/recode
Up: Video
Next: Multiplexing in the compressed domain
Previous Page: Hardware decode/mux/recode
Next Page: Multiplexing in the compressed domain

Software decode/mux/recode

This proposed path through the CMMC had not been implemented yet.

Clearly if a CMMC is to support a large number of conferences at a time, we cannot expect every remote user to have a hardware codec allocated to them. To this ends, we propose to use a compression/decompression engine derived from the INRIA IVS software to perform the functionality of a codec entirely in software. The CMMC is based around a SparcServer 690MP, which should eventually have four processors rated at 130MIPS, which should allow us to decode and encode a number of video streams simultaneously.

Three modes are envisaged for this path through the CMMC:

1. Several streams are decoded in software on the SparcServer. The decoded streams are mixed into one series of images, which are then recoded within the SparcServer.
2. As 1, except the decoded multiplexed image stream is converted to analog video , before being recompressed in a hardware codec. As decoding is generally less expensive than encoding, this will allow higher frame rates to be achieved than in 1, whilst still only requiring one hardware codec for encoding. It should be noted that the decompression part of this hardware codec is still available for participation in another conference using hardware multiplexing. In general, if multicast is used for output, more decoders are required that encoders, irrespective of whether they are hardware or software.
3. A single stream is decoded in software in the Server, and then converted to analog video for multiplexing into a hardware mux'ed conference. This allows the server to supplement a hardware mux'ed conference which is generally occurring at a quality level above what could be supported entirely in software.