[M4IF Technotes] R: About Transcoding

Lucia valentini coritel.it
Mon Jul 7 13:04:01 EDT 2003


Hello Jeongnam,
Thank you for your answer: it is precious to me, especially in deciding
which kind of architecture should be implemented in the transcoder. From
what i can see, a FPDT should be preferred to a DDT architecture....however,
to better justify a choise, i would like to read the paper about the author
you mentioned in your reply, could you please let me know his name?
Thank you very much.
Regards,
Lucia
> -----Messaggio originale-----
> Da: Jeongnam Youn [mailto:youn_98043   yahoo.com]
> Inviato: sabato 28 giugno 2003 1.51
> A: Lucia
> Oggetto: Re: About Transcoding
> Priorita: Alta
>
>
>
> Hello Lucia,
>
> Thanks for your interest in my work. The real-time
> transcoding actually depends on what size of picture
> you are dealing with. I don't have specific MIPS data.
> However, when I was in school(well, it's been passed 3
> years already), we implemented the transcoder. At that
> time, we were able to run it around 10 to 15
> frames/sec with QCIF size picture on 300 Mhz Pentimum
> if I remember the numbers correctly. So, you are
> considering QCIF format, I guess you will be able to
> do real-time transcoding. For CIF format, I am now
> working on JVT transcoding at 15 frames/sec and looks
> good on 2 Ghz Pentium too. However, you may need more
> optimization for faster processing. I prefer FPDT due
> to its flexibility. Some people claimed that DDT is
> faster FPDT before. However, I talked one of the
> famous authors in DDT and he agreed there was a
> mistake in computational complexity analysis in his
> paper.
>
> For the quantization error issue, I don't have the
> paper with me right now in my office. I will check it
> later and let you know.
>
> Thanks for your inquiry and hope you enjoyed the work
> in your project.
>
> Thanks,
>
> Regards,
>
> Jeongnam
>
> --- Lucia <valentini   coritel.it> wrote:
> >
> > >  Hi Jeongnam,
> > >
> > >  I'm a researcher of the university of Rome.
> > >
> > >  I've read some of your papers about video
> > transcoding; in
> > >  particular i'm very interested in bit-rate
> > reduction transcoding,
> > >  in fact the team i'm working with, wants to
> > realize a software
> > >  for transcoding that should work under linux in
> > real time.
> > >  Precisely we are focusing on bit-rate transcoding
> > during a video
> > >  conference. The scenario we are considering is
> > composed by 3 PCs,
> > >  representing the 2 terminals and the transcoding
> > gateway. We use
> > >  an MPEG4 coder with Simple Profile and only 1
> > object(Momusys
> > >  Reference Software). I wonder if the CPDT, FPDT,
> > DDT
> > >  architectures (that you presented in "Video
> > Transcoder
> > >  Architecture for Bit Rate Scaling of H.263 Bit
> > Streams") are fast
> > >  enough to be used in a test bed considering that
> > coding and
> > >  decoding are performed via software (with
> > Pentium4 at 2Ghz).
> > >  Regarding to the computational complexity of the
> > above cited
> > >  video transcoding architectures, do you have any
> > results ( in
> > >  MIPS) for the test sequences?
> > >
> > >  The last question is about the architecture's
> > pictures in the
> > >  aforementioned paper: I guess in fig.3 and fig.4
> > the signal to
> > >  calculate the quantization error, En, should be
> > taken after the
> > >  first inverse quantizer, and not after adding to
> > the inverse
> > >  quantized signal the MC error itself. Is it
> > right?
> > >
> > >  Thanks in advance.
> > >  Best Regards.
> > >
> > >  Lucia
> > >
> > >
> >
> _____________________________________________________________________
> > >  Lucia Valentini
> > >  CoRiTel-Consorzio di Ricerca sulle
> > Telecomunicazioni
> > >  URL     : http://www.coritel.it
> > >  E-mail  : valentini   coritel.it
> > >  Tel.    : +39 06 72582875
> > >  Fax.    : +39 06 72583002
> > >
> >
> ______________________________________________________________________
> > >
> > >
> > >
> > >
> > >
>
>
> __________________________________
> Do you Yahoo!?
> SBC Yahoo! DSL - Now only $29.95 per month!
> http://sbc.yahoo.com
>



More information about the Mp4-tech mailing list