Re: Issue-224 3D approach - disparity rather than (translation and condition)

From: Glenn Adams <<>> Date: Monday, 26 January 2015 14:59

On Mon, Jan 26, 2015 at 7:45 AM, Nigel Megitt <<>> wrote:
Thanks Glenn,

I like it (as you might have guessed!). A couple of follow-up points:

1. Do we still need "usesStereo" for anything?

it is still useful metadata

2. We should define that, relative to a value of 'zero', positive values should result in the image appearing further from the viewer than the plane of the display, and negative values should result in the image appearing closer to the viewer. This is ambiguous right now.

actually, i need to disallow negative values; the value is the horizontal disparity value, not depth; so it is up to author to choose desired disparity

Why do you need to disallow negative values? The solution doesn't work without them, or at best only defines that half of the range of depths can be supported, with the display being at one extreme of depth.

Plus a note to ourselves:

The disparity approach fits matches the request from SMPTE as well as what's specified by ETSI/DVB. Where those latter differ is that in the case of the SMPTE Digital Cinema spec the value is halved, and the left and right image each offset by that half-value; whereas in the case of DVB the value is applied equally both to the left and right image, resulting in double the apparent disparity compared to the same value in the SMPTE spec. As long as we're clear, which I think we are right now ("… is evenly divided along the horizontal axis between left and right stereoscopic images."), there's no problem.

Kind regards,


From: Glenn Adams <<>>
Date: Sunday, 25 January 2015 21:50
To: Nigel Megitt <<>>
Cc: TTWG <<>>
Subject: Re: Issue-224 3D approach - disparity rather than (translation and condition)

I have decided to submit to the apparent preference for specifying disparity directly, and have consequently changed from tts:translate to tts:disparity in [1].


On Tue, Jan 20, 2015 at 3:22 AM, Nigel Megitt <<>> wrote:

I see you have created update

to address issue-224, for 3D disparity. It looks as though the approach
you've taken is to allow the same document to be processed twice, once for
the left image and once for the right image for a stereoscopic display,
and to allow translation to be specified, being dependent on a parameter
and using the condition attribute.

Can I propose an alternate way to achieve stereoscopic object placement
that may be more amenable to simple, i.e. single pass, processing? This
would be to add a tts:disparity style attribute, whose value would be a
<length>, positive or negative. This would be inherited and animatable,
and apply to region, div or p (possibly a span too). Positive values imply
that the image is behind the plane of display and negative values imply
that the image is in front of the plane of display.

For example see [1] §4.2.1. Following the references, this seems to be how
it's done in DVB [2].

[1] ETSI TS 101 600 C1.1.1 (2012-05)

[2] ETSI EN 300 743 V1.4.1 (2011-10)


A good description from [2] (p. 34) is:

> Disparity is the difference between the horizontal positions of a pixel
>representing the same point in space in the right and left views of a
>plano-stereoscopic image. Positive disparity values move the subtitle
>objects enclosed by a subregion away from the viewer whilst negative
>values move them towards the viewer. A value of zero places the objects
>enclosed by that subregion in the plane of the display screen.

And from a little further down:

> A positive disparity shift value for example of +7 will result in a
>shift of 7 pixels to the left in the left subtitle subregion image and a
>shift of 7 pixels to the right in the right subtitle subregion image. A
>negative disparity shift value of -7 will result in a shift of 7 pixels
>to the right in the left subtitle subregion image and a shift of 7 pixels
>to the left in the right subtitle subregion image. Note that the actual
>disparity of the displayed subtitle is therefore double the value of the
>disparity shift values signalled in the disparity integer and/or
>fractional fields […]

Kind regards,


Received on Monday, 26 January 2015 15:05:44 UTC