[M4IF Technotes] AC prediction, follow document or follow reference software?

[Gary Sullivan]

Lefan, 
That appears to be a bug in the MPEG-4 visual spec.  I just looked too,
and 
I couldn't find a clear statement that non-intra macroblocks
are not used for AC coefficient prediction of Intra MBs, although
I think it would be silly to use them.
I don't recall whether that problem has been
previously reported to MPEG or not.  I will try to do so.
Best Regards,
Gary Sullivan
+> -----Original Message-----
+> From: Lefan Zhong [mailto:Lefan   mediaq.com] 
+> Sent: Wednesday, October 23, 2023 4:25 PM
+> To: technotes   lists.m4if.org
+> Subject: [M4IF Technotes] AC prediction, follow document or 
+> follow reference software?
+> 
+> 
+> Hello folks,
+> 
+> In MPEG-4 document 14496-2, subclause 7.4.3.3, adaptive ac 
+> coefficient prediction, it says:
+> 
+> If the prediction block (block 'A' or block 'C') is outside 
+> of the boundary of VOP or video packet, then all the 
+> prediction coefficients of that block are assumed to be zero.
+> 
+> 
+> Beware that statement doesn't tell us what to do when the 
+> prediction block is 'inter'. Should we use the prediction 
+> block coefficients to do prediction or set to zeros when the 
+> prediction block is 'inter' type?
+> 
+> In the comformance test bitstream Hit007.m4v, 1st P-VOP, the 
+> 2nd macroblock (coordinate (16,0)), the macroblock is an 
+> INTRA block, and AC prediction flag is on, and the 1st 
+> Y-block needs to do AC prediction from the left block. But 
+> left block is an INTER block, the AC coefficients are: 
+> 
+> -1 0 0 -1 0 0 0 0
+> 1 -1 0  0 0 0 0 0
+> 0 -1 0  0 0 0 0 0
+> 0  0 0  0 0 0 0 0
+> 0  0 0  0 0 0 0 0
+> 0  0 0  0 0 0 0 0
+> 0  0 0  0 0 0 0 0
+> 0  0 0  0 0 0 0 0
+> 
+> Should we use
+> 1024
+> 1
+> 0
+> 0
+> 0
+> 0
+> 0
+> 0
+> to do prediction or use
+> 1024
+> 0
+> 0
+> 0
+> 0
+> 0
+> 0
+> 0
+> to do prediction?
+> 
+> In the reference softwares (no matter momusys or microsoft, 
+> no matter version 1 or version 2), they all use zeros to do 
+> prediction in such cases.
+> 
+> For scence of prediction, it is better to use zeros in such 
+> cases. But that is conflict to document.
+> 
+> Thank you for your consideration.
+> 
+> Lefan
+> 
+> _______________________________________________
+> Technotes mailing list
+> Technotes   lists.m4if.org
+> http://lists.m4if.org/mailman/listinfo/technotes
+>