[M4IF Technotes] MVD

[Gary Sullivan]

Response in-line:
+> -----Original Message-----
+> From: cloverleafland   mail.hongkong.com 
+> [mailto:cloverleafland   mail.hongkong.com] 
+> Sent: Saturday, January 18, 2024 1:16 AM
+> To: technotes   lists.m4if.org
+> Subject: [M4IF Technotes] MVD
+> 
+> 
+> Dear all,
+> 
+> when viewing the standard 14496-2, I get confused with the 
+> vlc for MVD. There mentioned to have a flag called 
+> vop_fcode_forward and vop_fcode_backward to mark the range 
+> of MVD, but from the Table B-12 and Table B-29, both mention 
+> the range of MVD to be coded in [-16, 16]. However, in the 
+> vop_focde_forward/backward, the rnage of motion vector would 
+> be upto range [-2048,2047]. Would anyone explain why would 
+> the case be like this?
When the "vop_fcode_xyz" is not equal to 1, additional bits are sent
in "xyz_mv_residual" to specify the LSBs of the MV (where "xyz" is
some string of characters).
+> 
+> Also, for the MVD, does it mean to be the motion vector 
+> difference with the motion vector of the one previous block 
+> just before the current block?
It's the difference relative to a prediction of the value of the MV.
But the prediction is formed in a higher-efficiency way
than just taking the MV of the preceding MB.
+> 
+> Also, in the Table B-12 and Table B-29, why there is two 
+> table for the same purpose? Which one should be used instead?
Those two tables have different purposes.  I think one of them is
for ordinary video MVs with fractional-sample precision and the
other is for motion compensation of shapes with full-sample precision.
It might have been possible to use Table B-12 for both of them somehow,
but it isn't really a serious problem to have two tables (if that's
the biggest problem you have in implementing the spec, you're in a
class by yourself).
+> 
+> Thanks so much... 
+> 
+> Regards,
+> Clover  
+> ----------------------------------------------
+>  Åwªï¨Ï¥ÎHongKong.com¶l¥ó¨t²Î
+>  Thank you for using hongkong.com Email system
+> 
+> _______________________________________________
+> Technotes mailing list
+> Technotes   lists.m4if.org
+> http://lists.m4if.org/mailman/listinfo/technotes
+>