[FFmpeg-devel] DCT coeffs storing order

Germán Milano g.milano87 at gmail.com
Mon Nov 7 14:36:36 CET 2011


Hi,

Thanks again.

I think my questions are oriented to FFmpeg implementation mostly. In the
question

> When DCT coefficients are stored in the h->mb, in which order are they
> stored? I mean, do the first 256 correspond to luma, then the following
> correspond to one of the chromas, and the final 256 correspond to the
> second chroma component?. In case they're stored like i mentioned, which
is
> the order they appear in the array within each component?, e.g., the 256
> luma coeffs are stored in 'raster scan order'?

i was just asking how FFmpeg deals with the storing of the coeffs. It's
really important that i get this right, because i want to do a
requantification (for transRATING purpouses).

Thanks, and sorry i'm bothering you this much.

Jeremy

On Sun, Nov 6, 2011 at 2:15 PM, Michael Niedermayer <michaelni at gmx.at>wrote:

> On Sun, Nov 06, 2011 at 12:29:49PM -0200, Germán Milano wrote:
> > Hi everyone,
> > I was hoping someone could help me with the following questions:
> >
> > When DCT coefficients are stored in the h->mb, in which order are they
> > stored? I mean, do the first 256 correspond to luma, then the following
> > correspond to one of the chromas, and the final 256 correspond to the
> > second chroma component?. In case they're stored like i mentioned, which
> is
> > the order they appear in the array within each component?, e.g., the 256
> > luma coeffs are stored in 'raster scan order'?
> >
> >
> > Also, it seems h->mb_luma_dc stores values for luma and chroma, since it
> > depends of the *p* parameter, which refers to the plane. Is this right?
> If
> > yes, in what order are they stored?
> >
> >
> > I printed out the info for an "I" macroblock and this is what I obtained
> > (all other coefficients that don't appear have a velue of zero):
> >
> >  * CBP = 15
> >  * h -> mb_luma_dc[0][0] = 2
> >  * h -> mb[82] = -320
> >
> > I can't get any conclusions out of these values without knowing the
> answers
> > to my previous questions.. Any help will be truly appreciated.
>
> I suggest you force the whole mb to 0 and then set just 1 element to
> non zero and look at what happens to the picture. That way you will
> see what each coeff does
>
> Also dont forget that there is intra prediction
>
> [...]
> --
> Michael     GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB
>
> The greatest way to live with honor in this world is to be what we pretend
> to be. -- Socrates
>
> _______________________________________________
> ffmpeg-devel mailing list
> ffmpeg-devel at ffmpeg.org
> http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
>
>


More information about the ffmpeg-devel mailing list