[FFmpeg-devel] [PATCH] H.264 predictive lossless: again

[Michael Niedermayer]

On Sun, Nov 30, 2008 at 09:34:38PM +0100, Michael Niedermayer wrote:
> On Sun, Nov 30, 2008 at 12:24:49PM -0800, Jason Garrett-Glaser wrote:
> > > Also the add_pixels code could be replaced to do the down/right prediction
> > > of the residual on top of the already predicted block, this would avoid
> > > the prediction removal.
> > 
> > Again, the main problem here is the fact that in the same MB, in the
> > same function call, one can end up using both regular add_pixels and
> > predictive lossless.
> 
> iam fully aware of this
> 
> 
> >  That means that we cannot modify add_pixels to
> > do anything related to predictive lossless because then we can't
> > correctly decode either.
> 
> IIRC there was somethig like 
> idct4x4_add()
> if(transform_bypass && theprofile && (hor || ver)){
>     if(hor)
>         do_hori_predict_4x4_c()
>     else
>         do_vert_predict_4x4_c()
> }
> 
> just do
> 
> if(transform_bypass && theprofile && (hor || ver)){
>     if(hor)
>         do_hori_predict_4x4_c()
>     else
>         do_vert_predict_4x4_c()
> }else
>     idct4x4_add()
> 

i think i should elaborate a little more on this, my suggestion is of
course to change the do_hori_predict_4x4_c / do_vert_predict_4x4_c here to do
the pixel_add too. Somehow it seems iam not too good in clearly explaining
my thoughts today ...

[...]
-- 
Michael     GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB

Opposition brings concord. Out of discord comes the fairest harmony.
-- Heraclitus
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: Digital signature
URL: <http://lists.mplayerhq.hu/pipermail/ffmpeg-devel/attachments/20081130/80b60a16/attachment.pgp>