Re: Media queries for stereoscopic 3D displays?

On Sep 4, 2012, at 19:31 , Douglas Crosher <dtc-style@scieneer.com> wrote:

> On 09/05/2012 09:12 AM, David Singer wrote:
> ...
>> I think the question of whether the output surface is a single view,
>> stereo, or auto-stereo could be a CSS question, whereas the technology
>> to achieve it (e.g. dual or multi view, or single-stream frame packing)
>> should be out of scope.
> 
> An approach that can target many technologies would have advantages.
> 
> Note that image and video content could well be tagged with a declared
> format and converted to the target technology by the web browser and
> this matter seems outside the scope of the style forum.  However a
> website may have optimised content to target specific formats and it may
> still be handy to have media queries for these.  If such media queries
> were used to select among un-tagged content then technologies not
> handled by media queries would not be supported (unless the format could
> be inferred from the CSS media query which would seem an abuse of CSS).
> 
> For the purpose of styling, a general target would mean supporting a
> fluid range of perspectives from left to right.  This could be done
> explicitly with a parameter ranging from -1.0 to 1.0 for left to right,
> or implicitly by allowing the depth of styled surfaces to be defined.
> 
> A proposal to add depth to styling would be ambitions but would seem the
> most general solution and perhaps the best investment, and it would also
> lead to the most consistent presentation.  Even on single view surfaces
> the depth could be used to support automatic perspective, highlights,
> shadows, focus, panning, and wiggle, cues.
> 
> The question becomes how to allow the depth to be specified in CSS?
> 
> Some options:
> * Just the corners.
> * Plus bump maps.
> * Bezier or NURBS surfaces.
> * A calculation.
> 
> Some other issues to consider:
> * Viewpoint
> * Lighting
> * Surface properties

Right.  some things need to be 'styled in depth' (notably captions and sub-titles) but this is complex and temporal.  The reason is two-fold -- you want the captions to be at the same apparent convergence as the main part of the scene, yet you don't want them 'punching holes' in things that appear to be closer.

Even if we were to style HTML in 3D, we'd need to look at this -- not only 3D transforms, but simple depth positioning, and so on.

The simple way to do it is to either specify the depth and let the system work out the disparity (the left/right eye shift) or specify the disparity directly.  I suspect this will rapidly come into collision with the difficulties with physical measurements, as some of the calculations concern eye-eye distance and distance to the image plane (screen etc.), both of which are true physical questions, yet often unknown to CSS.

> 
>> But we need to decide what the semantics of a media query are (maybe we
>> already did, as there have been previous discussions).  When do they
>> overlap, and so on?
> 
> It would be good to review any prior discussion and would you happen to
> recall where the previous discussion occurred and know any keywords to
> locate it?


I am collating scraps from conversations at Blu-ray, MPEG, 3GPP, and so on!!

David Singer
Multimedia and Software Standards, Apple Inc.

Received on Wednesday, 5 September 2012 23:26:35 UTC