[M4IF Technotes] h.263 motion vector predictor

[Ariel David]

Yes,
there are GOB headers in the stream.
So I understand that a change of GOB is like a change of VOP .  
Prediction cannot be made from one GOB to another just as it can not be made from one VOP to another.
Is this correct?
Ariel
-----Original Message-----
From: Gary Sullivan [mailto:garysull   windows.microsoft.com]
Sent: Saturday, May 24, 2024 11:56 PM
To: Ariel David
Subject: RE: [M4IF Technotes] h.263 motion vector predictor
Does that stream have GOB headers in it?  I'm trying to remember whether
the presence of GOB headers cuts the prediction of MVs from the MBs in
the row above.  A quick look in either the H.263 or MPEG-4 spec should
answer that.
-Gary
+> -----Original Message-----
+> From: Ariel David [mailto:Ariel.David   ParthusCeva.com] 
+> Sent: Thursday, May 22, 2024 12:27 AM
+> To: Gary Sullivan
+> Subject: RE: [M4IF Technotes] h.263 motion vector predictor
+> 
+> 
+> Hi,
+> I'm sorry to nag about this issue, but there is something I 
+> still don't understand.
+> 
+> I am currently working on decoding the test stream 
+> "vcon-ge-16-L1.m4v".  This is a short header video stream.
+> When I decode the motion vectors using the median predictor 
+> the output picture comes out bad.
+> When I decode the motion vectors using the left macroblock 
+> only as the predictor the picture comes out OK.
+> This is the reason for my questions about the subject.  
+> There is simply something that I don't understand here.
+> 
+> As far as I can see there can a number of answers:
+> 1.  The stream was not encoded correctly in that the 
+> predictors used in the encoder are not median.
+> 2.  This is some special case of prediction that I am not aware of.
+> 
+> Thanks for the previous answers,
+> 
+> Ariel
+> 
+> 
+> 
+> 
+> -----Original Message-----
+> From: Gary Sullivan [mailto:garysull   windows.microsoft.com]
+> Sent: Wednesday, May 21, 2024 7:23 PM
+> To: Ariel David; technotes   lists.m4if.org
+> Subject: RE: [M4IF Technotes] h.263 motion vector predictor
+> 
+> 
+> Skipped MBs in P pictures use (0,0) MVs.  Doesn't matter 
+> whether short
+> header or not.
+> 
+> -Gary S.
+> 
+> +> -----Original Message-----
+> +> From: technotes-admin   lists.m4if.org 
+> +> [mailto:technotes-admin   lists.m4if.org] On Behalf Of Ariel David
+> +> Sent: Wednesday, May 21, 2024 8:25 AM
+> +> To: technotes   lists.m4if.org
+> +> Subject: RE: [M4IF Technotes] h.263 motion vector predictor
+> +> 
+> +> 
+> +> Maybe the problem is in my median prediction.
+> +> 
+> +> are predictors that are skipped macroblock computed 
+> +> differently or are they considered 0,0 motion vector.
+> +> For example:
+> +> if the top left MV is 0,0, the top MV is 1,1 and the left is 
+> +> skipped will the predictor be 0,0 (regular median) or 1,1 
+> +> (since the left is not considered).
+> +> 
+> +> Ariel
+> +> 
+> +> -----Original Message-----
+> +> From: Herbert Thoma [mailto:tma   iis.fhg.de]
+> +> Sent: Wednesday, May 21, 2024 5:39 PM
+> +> To: Ariel David
+> +> Cc: technotes   lists.m4if.org
+> +> Subject: Re: [M4IF Technotes] h.263 motion vector predictor
+> +> 
+> +> 
+> +> Ariel David schrieb:
+> +> > 
+> +> > Hi,
+> +> > 
+> +> > I am trying to decode mpeg-4 video files with short header 
+> +> (h.263 format).
+> +> > I've looked though the mpeg-4 standard and found not 
+> +> mention of a different method for motion vector predictor.
+> +> > The mpeg-4 method is the median of the 3 adjacent macroblocks.
+> +> > This does not seem to work in h.263 (mpeg-4 with short headers).
+> +> 
+> +> Why not? It's exactly the same in H.263.
+> +> 
+> +>  Herbert.
+> +> 
+> +> > I assume it's more like mpeg-2, either left or top 
+> +> macroblock, but I can't find any mention in the standard.
+> +> > 
+> +> > Can someone help me out?
+> +> > 
+> +> > Thanks,
+> +> > 
+> +> > Ariel
+> +> > 
+> +> > _______________________________________________
+> +> > Technotes mailing list
+> +> > Technotes   lists.m4if.org
+> +> > http://lists.m4if.org/mailman/listinfo/technotes
+> +> 
+> +> -- 
+> +> Herbert Thoma
+> +> FhG-IIS A, Studio Department
+> +> Am Wolfsmantel 33, 91058 Erlangen, Germany
+> +> Phone: +49-9131-776-323
+> +> Fax:   +49-9131-776-399
+> +> email: tma   iis.fhg.de
+> +> www: http://www.iis.fhg.de/
+> +> 
+> +> 
+> +> _______________________________________________
+> +> Technotes mailing list
+> +> Technotes   lists.m4if.org
+> +> http://lists.m4if.org/mailman/listinfo/technotes
+> +> 
+> 
+>