Re: [GnomeMeeting-list] OpenH323 + Camera + H.263 card support



Am Freitag, 11. Februar 2005 18:07 schrieb Kilian Krause:
> Hi Sharma,
>
> > Also somwhere I read that MPEG-4 is same as H.263. So does that mean
> > that I can use MPEG-4 card for having H.263 support.
>
> MPEG-4 is a general video description in terms of standards. DivX and
> XviD are the non-streaming versions, H.264 is the streaming version of
> MPEG-4. H.263 is the streaming version of MPEG-2 and H.261 of MPEG-1.
> (please Craig correct me if i'm wrong)

Mpeg-4 Part 2 is also known as DivX and Xvid. Although Mpeg-4 offers many 
things Mpeg-2 does not, especially in terms of higher level object 
representations, but in reality they don't differ much. The better coding 
efficiency of Mpeg-4 is only due to some improvements in the details of the 
coding scheme. Mpeg-4 has a higher resolution for the motion compensation, 
offers different sizes for the motion compensation blocks and some other 
niceties. H.263 is more or less the same as Mpeg-2, H.263+ is more or less 
the same as Mpeg-4 Part 2.

Mpeg-4 Part 10 is also known as AVC or H.264 (in contrast to the older 
standards, MPEG and ITU have created a joint project and the standards are 
this time not only similar, but exactly the same). Part 10 is based on the 
works of Part 2 and h.263+, so again it is based on the same coding 
principles (DCT and Motion Compensation), but it offers again more 
flexibility in the details of coding. It introduces some new ideas like CABAC 
instead of Huffman coding, which seems to be very efficient ...

> That's about the same discussion as why Ogg Speex (streaming format)
> isn't Ogg Vorbis (non-streaming format). They use the same ideas of
> codec design, but have different targets in terms of stream loss and
> connection reliability vs. compactation and image quality.

Here you are completely wrong. Speex is speech coding. The coding design is 
based on a model of speech synthesis - The oscilation of the vocal chord has 
a fundamental frequency, this fundamental frequency is transmitted, the 
higher frequencies of the vocal chord are formed by the oral apparatus, which 
can be thought of as a filter system, whose coefficients are transferred, 
too. These principles are used by every speech codec in use.

Vorbis (as mp3) are modelled after the human aural system. Quiet tones are 
masked by louder tones of a similar frequency, so you have to transfer only 
the louder tone.

This difference is the reason for music transferred over a speech codec often 
sounding bad. If the music is similar to spoken language (A solo vocalist, a 
single trumpet or even a violin), it sounds quite good, but transferring a 
drum or a full band or orchestra wont  work well.

If a codec is suited for streaming depends most times not on the codec itself, 
but on the container format (.avi, .mov, .ogg, .mp4). Critical for realtime 
streaming (Audio/Video conference, Live TV) are coding delays introduced by 
the codec.

Hope this clarifys the differences,

Greetings,

Stefan

-- 
Stefan Brüns  /  Kastanienweg 6 - Zimmer 1206  /  52074 Aachen
mailto:lurch gmx li  http://www.kawo1.rwth-aachen.de/~lurchi/
   phone: +49 241 169-4206     mobile: +49 160 3797725

Attachment: pgpyGpreAYcEL.pgp
Description: PGP signature



[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]