[M4IF Technotes] Chrominance positions

[the_ether]

Thanks for the reply.
Certainly F2 appears to say that a VOP can only ever be positioned at even
co-ordinates and therefore positions such as (3,2) are no-nos. I'm surprised
as this would have quite a heavy restriction, also on things like BIFS.
I reckon that F2 is for guidance not a rule and that odd-numbered
co-ordinates are possible. Having said that I wonder what the decoder would
think of all that as you would have conflicts in the calculation of the
pixel value to display between the chrominance elements to the left of the
luminance pixel and the 'odd numbered' ones to the right.
Thinking further, the bounding region may be fixed to only even-numbered
positions but the masked image could fall anywhere. The VOP position is
simply just a bounding region ad will contain non object pixels as well as
the 'desired' ones.
I'll go with the view that a VOP has to be on an even-numbered position and
so the chrominance values ca be calculated in one sweep of the screen
without regard to VOPs.
g.
----- Original Message -----
From: "Arcin Bozkurt" <arcin   lumictech.com>
To: "the_ether" <the_ether   btinternet.com>
Sent: 14 December 2023 11:07
Subject: Re: [M4IF Technotes] Chrominance positions
> On Thursday 13 December 2023 23:11, you wrote:
> > Are the positions of the chrominance pixels in fixed positions on the
> > screen or are they relative to the positions of VOPs.
> >
> > That is to say, the normal description is
> >
> > Y  Y  Y  Y  Y  Y
> >   c       c       c
> > Y  Y  Y  Y  Y  Y
> >
> > ie one Cb and Cr per 4 Y pixels
> >
> > However, what if the VOP sits at position (1,0)? I presume that we would
> > now have
> >
> > Y  Y  Y  Y  Y  Y
> >       C     C
> > Y  Y  Y  Y  Y  Y
>
> as far as i know, you cannot have a VOP that sits at an odd numbered
topleft
> corner... Or maybe I should say, you can, but as you said, you will have
to
> recalculate new set of C values from the originals.
>
> section 6.1.3.6 talks about the 420 format and refers to section F.2 for

> formation of VOP. In this F2, the tightest bouding rectangle around the
VOP
> is chosen starting from an even numbered topleft corner.
>
>
> > ie the C values would have to be re-calculated as they now relate to
> > different Y pixels.
> >
> > So I assume  that the chrominance values move relative to the VOP. In
other
> > words I cannot blindly calculate the YCbCr values for the whole screen.
I
> > have to decide on the VOP positions before calculating Cb, Cr.
> >
> > Is this correct?
> >
> > Regards
> >
> > graham
> >
> >
> > _______________________________________________
> > Technotes mailing list
> > Technotes   lists.m4if.org
> > http://lists.m4if.org/mailman/listinfo/technotes