[M4IF Technotes] (video_object_layer_width / height) not a mu ltiple of 16

[Gary Sullivan]

Response below in line:
+> -----Original Message-----
+> From: Robert Beattie [mailto:r.beattie   indigovision.com] 
+> Sent: Wednesday, May 22, 2024 4:42 AM
+> To: 'Biju Ravindran'; 'technotes   lists.m4if.org'
+> Subject: RE: [M4IF Technotes] (video_object_layer_width / 
+> height) not a mu ltiple of 16
+> 
+> 
+> >    If we do motion estimation on outside the boundary
+> > after extending the edge pels, it can give a match
+> > which will result in reduced bit rate. So there is no
+> > need of restricting the motion vector if the edge pels
+> > are copied outside the effective area.
+> 
+> 
+> This is true Biju. What I am saying is that edge pels should 
+> be extended
+> from line 241 rather than line 255 for the image resolution 
+> described. 
No.
+> 
+> My understanding of the standard is that the decoder can not 
+> treat a 352x256
+> and 352x242 in the same way when extending the boundary.
That is not the right interpretation.
There is no difference between the decoding process
for a 352x256 image and and for a 352x242 image (assuming no shape
coding is used).  The only difference is that the final output of
the decoding process is cropped before being sent to the (non-
normative) display.  Any motion compensation over
picture boundaries operates across the bottom edge by
extrapolating row 255 in both cases, not row 241.
The padding process is based on the bounding rectangle (at
least in the absence of shape coding).
The bounding rectangle is the vop_width and vop_height
extended to each be a multiple of 16.  The vop_width
and vop_height do not affect the decoding process
except to determine how many macroblocks to decode.
You are not the only one who has been tripped up by
the way this is specified.  It should have been stated
more clearly.  (I think this is already listed in the
MPEG problem report list, but I'll try to submit it
to the next meeting to make sure.)
+> 
+> I have a concern with this as our (vhdl / fpga) decoder 
+> successfully passed
+> the compliance tests even though it did not (in my 
+> understanding) support
+> this case correctly. 
Perhaps the conformance test streams did not test this case.
+> 
+> This being the case is the coverage of the conformance streams really
+> adequate ? There are several other similar cases which cause 
+> me concern. I
+> would be interested if anyone else has found possible holes in the
+> conformance tests ?
The conformance tests are not claimed to be exhaustive.  They are
simply the best that people managed to volunteer to contribute.
Best Regards,
Gary Sullivan
+> 
+> Your's waiting to be corrected,
+> 
+> Dr Rob Beattie
+> Indigo
+> 
+> 
+> 
+>  
+> The information in this e-mail is confidential and for use by the
+> addressee(s) only. If you are not the intended recipient (or 
+> responsible for
+> delivery of the message to the intended recipient) please notify us
+> immediately on the above detailed phone number and delete 
+> the message from
+> your computer: you may not copy or forward this e-mail, or 
+> use or disclose
+> its contents to any other person. We thank you in 
+> anticipation for your
+> assistance. As internet communications are capable of data 
+> corruption no
+> responsibility is accepted for changes made to this message 
+> after it was
+> sent. For this reason it may be inappropriate to rely on information
+> contained in this e-mail without obtaining written 
+> confirmation of it. In
+> addition, no liability or responsibility is accepted for 
+> viruses and it is
+> your responsibility to scan any attachments to this e-mail. 
+> Nothing in this
+> e-mail shall constitute or be construed as constituting an 
+> offer, obligation
+> or an acceptance of any offer previously made. Opinions, 
+> comments and other
+> information in this e-mail that do not relate to the business of
+> IndigoVision Group plc, IndigoVision Limited and/or 
+> IndigoVision, Inc. shall
+> be understood as neither given nor endorsed by the companies 
+> or any of them.
+> _______________________________________________
+> Technotes mailing list
+> Technotes   lists.m4if.org
+> http://lists.m4if.org/mailman/listinfo/technotes
+>