[x264-devel] Re: [PATCH] set constraint_set[01]_flag when possible
Loïc Le Loarer
lll+vlc at m4x.org
Fri May 12 11:36:11 CEST 2006
Le Friday 12 May 2006 à 01:18:20 -0700, Loren Merritt a écrit:
> On Fri, 12 May 2006, Loïc Le Loarer wrote:
> >Le Thursday 11 May 2006 à 16:24:26 -0700, Loren Merritt a écrit:
> >
> >>I suppose you could argue that there is some value in setting
> >>constraint_set1_flag for Baseline files, because there are decoders that
> >>support Main but not all of Baseline.
> >
> >This is what I wanted to add. Because of this and because it is faster
> >to do deblocking on the fly as it is possible in MAIN.
>
> I don't buy speed as a reason. If I were implementing a decoder that
> supported both strategies of deblocking, I would select which one to use
> based on whether slicegroups or ASO were used in any given frame, not on
> the profile or constraint_set.
I agree for slicegroups but not for ASO, you cannot know if ASO is used
without parsing all the slice headers of a picture, it is not always
practical and not all decoders will do this.
> >By the way, by checking that parameter checks are really duplicated, I
> >discover that the profile can be set to BASELINE while to shouldn't:
> >when weighted_bipred_idc (b_weighted_bipred is x264) is not 0.
>
> weighted_bipred applies only to B-frames. If there are B-frames, it's
> already not Baseline. However, I did fail to forbid weighted_bipred when
> not using B-frames. Fixed that instead.
You are right, this is better.
On this point, the H.264 norm is a bit other restrictive,
weighted_bipred_idc has no meaning in Baseline, it shouldn't be
restricted like the direct_8x8_inference_flag.
Thank you.
--
Loïc
"heaven is not a place, it's a feeling"
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: Digital signature
Url : http://mailman.videolan.org/pipermail/x264-devel/attachments/20060512/b6f1022d/attachment.pgp
More information about the x264-devel
mailing list