[M4IF Technotes] About the "validity" in BME

Dhinesh dhinesh tataelxsi.co.in
Thu Jul 10 12:48:25 EDT 2003


Hi,
Refer to Section "7.5.2.3 Motion vector decoding"
in "ISO/IEC 14496-2:2001(E)".
The candidates order is MVs1, MVs2, MVs3, MVt1,
MVt2, MVt3.
If any of the candidates is valid (the first valid
one in the order),
its motion-vector is taken to be the
motion-vector-predictor.
If none of the candidates are valid,.
the motion-vector-predictor is set to zero.
The validity of the candidates is as defined
below:
The rules for MVs (MV-Shape)
-The MVs of the candidate MB with following
bab_type is *not valid*.
--bab_type = 2 (all_0)
--bab_type = 3 (all_255)
--bab_type = 4 (intra)
The rules for MVt (MV-Texture)
-The MVt of the candidate MB is *not valid* if
--the candidate MB is Transparent or
--the candidate is not in the current video-packet
or outside the bounding-rectangle.
Note : If the candidate-MB is semi-transparent and
4MV (MVt),
the texture motion-vectors of the
transparent-blocks are motion-vector-padded,
and will be used in prediction.
Regards,
Dhinesh
> Hi,everyone,
>  I get a question in shape coding BME.
>  In MPEG-4 VM18, when finding the MPVs,
> "by looking into MVs1,MVs2,MVs3,MV1,MV2,MV3 in
this order,
> MVPs is determined by taking the first
encountered MV that is valid".
> But what is "valid"? Suppose the neighborhood
BAB is all_0 or all_255 (i.e. type 3 or 4),
> they have no MVs, are they  valid? What's the
value of MVs if valid?
> Then how about the case of type 4(i.e.
intraCAE)?
>
> Thanks in advance.
> Z.G.F.



More information about the Mp4-tech mailing list