[vorbis-dev] Ogg VEO Codec, video in an Ogg stream

[Jelle Foks]

Sean Wieland wrote:
> 
> "Manuel Amador (Rudd-O)" wrote:
> >
> > On searchable databases:
> >
> > The data searching technologies are becoming so important that even
> > Oracle rolled out a database product called Internet file system, acting
> > as a hierarchical storage with searching capabilities.
> >
> > The thing is, every day we have to deal with an increasing amount of
> > data.  Even we geeks are noticing that it's becoming more and more
> > difficult to find what we want in the sea of information, never mind
> > mere mortals.  What will save us when later on we don't have multimedia
> > searching capabilities too?
> 
> My point was that database stuff is a job for Oracle/xSQL/whatever.  You
> want content description?  Try XML.  "Every tool has a purpose, and
> every purpose has a tool."  Searchable databases are for database
> software to deal with, not a multimedia standard.  A multimedia standard
> should just define the algorithms and processes required to encode and
> decode compliant streams, with the idea that newer standards offer
> higher quality and at lower bitrates than the previous ones.  I'm not
> saying that innovation should be stiffled either, far from it.  If
> there's a problem organizing information and finding what's relevant,
> then I'm all for finding a solution to that problem.  But including it
> in a multimedia standard is inappropriately low-level in my opinion.

I agree with this one here.

Everybody seems to forget in this discussion that the objects can help
to allow higher compression rates. Not every object has the same motion
properties, and the wireframe motion vectors and variation of coding
parameters between objects allow for pretty efficient motion
compensation and residue coding _if_ you have a good encoder. That's
where the objects in MPEG4 can actually be useful (plus for things such
as mixing of streams, multipoint transmission, etc.).

But, when you design an encoder that uses the objects for that purpose,
then the database will _not_ know that this round black-and-white object
is really called 'a soccer ball' until it does some object recognition
of it's own, and has some sort of lexicon for property to name mapping.
I think it's not correct to say that allowing objects in the multimedia
stream will make sure that we can create good databases that don't need
to read, compute, and index multimedia files so that we can search the
multimedia files. Even when you want to search for 'comparable objects',
you still need to scan the files and study the behaviour, movement,
shape and shape variations of the objects to be able to classify the
objects as 'similar'. Additionally, with MPEG4, no guarantee at all
exists that different brands of encoders will find similar objects and
encode them with similar properties as MPEG4 objects, so to create a
good index, your database will have to do it's own image segmentation
and object recognition to compensate for the variations between the
encoder implementations.

Concluding, this 'objects for databases sceme' will work well for
databases in which all streams are encoded with the same encoder, which
has enough quality to recognize and process all similar objects in the
streams in similar ways, or where the database performs detailed
processing of it's own.

If a multimedia stream is to be indexed, a description should be
generated and included at the source, it the form best suitable for
content description. XML sounds like a good, well defined, open standard
format for that.

Jelle.

> -Sean
> 
> --
>  +-----=[export-a-crypto-system-sig RSA-3-lines-PERL]=-----+
>  |#!/bin/perl -sp0777i<X+d*lMLa^*lN%0]dsXx++lMlN/dsM0<j]dsj|
>  |$/=unpack('H*',$_);$_=`echo 16dio\U$k"SK$/SM$n\EsN0p[lN*1|
>  |lK[d2%Sa2/d0$^Ixp"|dc`;s/\W//g;$_=pack('H*',/((..)*)$/)  |
>  +--------=[http://www.cypherspace.org/~adam/rsa/]=--------+
> "Most people would rather die than think, and most people do."
>                                        -- Bertrand Russell
> 
> --- >8 ----
> List archives:  http://www.xiph.org/archives/
> Ogg project homepage: http://www.xiph.org/ogg/
> To unsubscribe from this list, send a message to 'vorbis-dev-request at xiph.org'
> containing only the word 'unsubscribe' in the body.  No subject is needed.
> Unsubscribe messages sent to the list will be ignored/filtered.

--- >8 ----
List archives:  http://www.xiph.org/archives/
Ogg project homepage: http://www.xiph.org/ogg/
To unsubscribe from this list, send a message to 'vorbis-dev-request at xiph.org'
containing only the word 'unsubscribe' in the body.  No subject is needed.
Unsubscribe messages sent to the list will be ignored/filtered.