[XviD-devel] VHQ

Christoph Lampert chl at math.uni-bonn.de
Wed Feb 12 12:35:49 CET 2003



Hi Radek,

that's great news! 

On Wed, 12 Feb 2003, Radek Czyz wrote:

> Heya everybody,
> 
> I have good news - I fixed all known VHQ bugs and I'm preparing to
> commit the code. Anyone would like to comment? What I mean is if
> someone would like to stop me for whatever reason, please do it now.
> 
> The code doesn't work with GMC yet (mcsel comparsion would not work)
> but I'm going to fix this soon. The result will be a perfect,
> brute force S_VOP/P_VOP selection and mcsel selection.
> 
> A bit about the flags which will appear in xvid.h:
> 
> global flag XVID_MODEDECISION_DCT will activate the code and use it
> for mode decision.
>  
> motion flags:
> QUARTERPELREFINE16_DCT - quarterpel refinement for 16x16 mode will be
>   performed using DCT (well in fact not dct, but BITS) comparsion. It
>   doesn't mean it will not be performed using SAD, so do not deactivate
>   PMV_QUARTERPELREFINE16 or suffer some PSNR degradation. I'll write why
>   this happens if someone asks me.

I didn't get that. Qpel refine is done how exactly? SAD _and_ BITS ?
 
> [...]

If you really implemented BITS instead of DCT, the "DCT"-names I proposed
are of course misleading. Why not call all the flags "XXX_BITS" instead of
"XXX_DCT"? 

Btw., I didn't even know there were that many bits left in our flags 
vector... Maybe we have to switch to 64-bit because of too many
features. ;-) 


gruel 




More information about the XviD-devel mailing list