[M4IF Technotes] Interpretation of vop_coded

[Kris Huber]

Dear Yogender,
The 2001 edition of the video spec., at least, describes the behavior you
were expecting (i.e., if rectangular then "the luminance and chrominance
planes of the reconstructed VOP shall be filled with the forward reference
VOP..." - look under vop_coded in subclause 6.3.5).  Filling a vop_coded=0
rectangular VOP with 128's is not standard, and is a bug in the reference
software if it is happening.  Even if vop_coding_type=0 (I-VOP), I think
vop_coded=0 in the header causes  a copy of the most recent VOP in the past
for which vop_coded=1 to be displayed.  Anything other than that is
unspecified by the standard.
As for when an encoder should use vop_coded=0 in the rectangular VOP case, I
notice that it allows you to effectively drop a frame even if
fixed_vop_rate=1 in the VOL header.  This might be necessary or convenient
to use for meeting profile complexity constraints.  It is a more efficient
way of skipping the coding of a VOP than constructing a P-frame consisting
entirely of not-coded macroblocks.  Alternatively, if fixed_vop_rate=0 is
used, one can still produce a sequence of VOPs with a  fixed VOP rate, and
not have to insert the (small) vop_coded=0 headers.  In some systems where
MPEG-4 video is used (including mp4 files I believe) the time-code
information in the VOP headers is redundant anyway, which is the only
information that comes in the vop_coded=0 VOP header.
Regards,
Kris Huber
-----Original Message-----
From: Yogendra Kumar Gupta <ykgupta   noida.hcltech.com>
To: technotes   lists.m4if.org
Date: Mon, 11 Feb 2024 19:40:19 +0530
Subject: [M4IF Technotes] Interpretation of vop_coded
Dear All,
Can anyone please explain the interpretation of the flag vop_coded (when
vop_coded = 0) when video_object_layer_shape == RECTANGULAR. The standard
gives details of the construction of reconstructed frame but does not give
any interpretation regarding the construction of the current frame (that
would be used for display). The Microsoft reference code puts a grey image
of colour 128 in the o/p which doesn't look very nice in between while
watching a video. One of the conformance bitstreams contains frames where
every alternate frame the flag vop_coded is set to 0 and the o/p looks like
a blinking one where you get a decoded image and a gray image alternatively.
Could a better solution be use the previous frame itself for display when
vop_coded = 0. Also I would like to know when the flag vop_coded needs to be
set to 0 on the encoding side.
Regards
Yogender Kumar Gupta