[XviD-devel] EDGE_SIZE == 64

[Christoph Lampert]

On Sun, 9 Mar 2003, Jim Hauxwell wrote:

> 
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
> 
> Gruel,
> 
> Check out table B-12 in the spec.
> 
> its for the out of bounds motion vectors.  It makes the calculation easier.
> The edge size is 64 as we can go -16 to +16 in half pel units therefore
> needing 32bytes at each side of the image of replicated edge data.

I know about motion vectors, but as I said we just need 16 for that, not
64. Halfpel units don't count, because we have 4 images for interpolation
modes, every one of the uses fullpel resolution.
-16 / +16 in fullpel is already covered by 16 pixels margins on the
left (for the minus) and 16 pixels on the right (for the plus). 

gruel 


> > -----Original Message-----
> > From: xvid-devel-bounces at xvid.org [mailto:xvid-devel-bounces at xvid.org]On
> > Behalf Of Christoph Lampert
> > Sent: 09 March 2003 16:45
> > To: xvid-devel at xvid.org
> > Subject: [XviD-devel] EDGE_SIZE == 64
> >
> >
> >
> > Hi,
> >
> > I'm sure this was asked before, but: Why is EDGE_SIZE defined
> > as 64 in image/image.h? That's far beyond what is needed for
> > unrestricted motion vectors. 16 should be sufficient for this, or not?
> >
> > If it's just to align lumi and chroma at same stride, then at least the
> > extra pixels should not be processed during image_setedges.
> >
> > Second question: Image interpolate has
> >
> > const uint32_t offset = EDGE_SIZE2 * (edged_width + 1);
> > // we only interpolate half of the edge area
> >
> > Why do we interpolate edged area at all? In Halfpel mode, two of four
> > edges will be constant anyway (vertical at top and bottom, horizontal at
> > left and right) and the others will just contain copies of the boundary
> > row/column again, so a memcpy call (like setedges) would be sufficient.
> >
> > gruel
> >
> > _______________________________________________
> > XviD-devel mailing list
> > XviD-devel at xvid.org
> > http://list.xvid.org/mailman/listinfo/xvid-devel
> -----BEGIN PGP SIGNATURE-----
> Version: PGP 8.0
> 
> iQA/AwUBPmuyDypQmIptkMWREQKiBQCfQDeuQvQdSwY6lgD1gXIxu+u7WvsAnRl9
> FJ6iluKgXaJbRiqKEDdUQdZR
> =tnWv
> -----END PGP SIGNATURE-----
> 
> _______________________________________________
> XviD-devel mailing list
> XviD-devel at xvid.org
> http://list.xvid.org/mailman/listinfo/xvid-devel
>