[M4IF Technotes] Independent Segment Coding

[Gary Sullivan]

Stathis is correct.  But the term "Independent Segment Decoding" (ISD)
as used in H.263 means more than that.  What both Stathis and Lucia
described is simply slice-structured coding.  I believe all
slice-structured video coding standards operate that way - in fact, even
GOB-structured video coding standards operate that way.
But independent segment decoding as specified in H.263 Annex R is
something more than that.
Slice structured decoding makes the decoding of each region independent
(or nearly independent) of the decoding of any other region of the same
decoded picture.
Independent segment decoding makes the decoding of each region
independent of the decoding of any other region of the same decoded
picture AND independent of any other region of the reference picture(s).
It treats the boundaries of the region being decoded like they were the
boundaries of the picture.
I believe the way to use MPEG-4 part 2 to achieve something similar to
H.263 ISD is to send each independent region as a separate "OBJECT" - as
a separate elementary stream.
Best Regards,
Gary Sullivan
+> -----Original Message-----
+> From: technotes-admin   lists.m4if.org 
+> [mailto:technotes-admin   lists.m4if.org] On Behalf Of Stathis 
+> Voukelatos
+> Sent: Monday, April 07, 2024 3:40 AM
+> To: Technotes
+> Subject: RE: [M4IF Technotes] Independent Segment Coding
+> 
+> 
+> Hi Lucia,
+> 
+> In MPEG-4 this is implemented by the video packet error 
+> resilience tool. AC/DC prediction and motion vector prediction
+> are reset on video packet boundaries, in the same way as in frame
+> boundaries. 
+> In the DC prediction case for example, if one of 
+> neighbouring DCT blocks
+> used to form the prediction does not belong to the same 
+> video packet as the
+> current block then its DC value is not used, but instead the value
+> of 1024 is used in the prediction process (which is the DC value for
+> mid-grey).
+> 
+> Stathis
+> 
+> 
+> Stathis Voukelatos, PhD      
+> Software Engineer       
+> IndigoVision Ltd     
+> The Edinburgh Technopole
+> Bush Loan, Edinburgh
+> Scotland, UK
+> EH26 0PJ
+> +44 (0)131 4757345      
+> +44 (0)131 4757201 (Fax)  
+> http://www.indigovision.com
+> 
+>  
+> 
+> > -----Original Message-----
+> > From: Lucia [mailto:valentini   coritel.it]
+> > Sent: Monday, April 07, 2024 11:02 AM
+> > To: Technotes
+> > Subject: [M4IF Technotes] Independent Segment Coding
+> > 
+> > 
+> > 
+> > Hi everybody,
+> > 
+> > does anybody know if in MPEG4 it is implemented 
+> > the "independent segment coding" as in H263?
+> > (i mean confining spatial prediction by not predicting DCT coeff.
+> > and Motion vectors from adiacent Macroblocks in a frame, 
+> but sending 
+> > their absolute value)
+> > 
+> > Thanks in advantage
+> > Lucia
+> > _______________________________________________
+> > Technotes mailing list
+> > Technotes   lists.m4if.org
+> > http://lists.m4if.org/mailman/listinfo/technotes
+> > 
+> _______________________________________________
+> Technotes mailing list
+> Technotes   lists.m4if.org
+> http://lists.m4if.org/mailman/listinfo/technotes
+>