Re: Extensions for stereoscopic 3D images?

Thank you for the link.  The Lytro appears to use an array of small lenses to capture a light field, and while there are some
similarities to stereoscopic images the processing is probably quite different and would warrant a separate schema.

If you have a left and right image, or a left to right sweep, then it is a relatively small step to extract a left and right pair to
present a stereoscopic 3D image.  Whereas the processing involved in converting the image obtained from a grid of lenses to a
picture is something quite different.

There may be some similarities.  Stereoscopic images have a 'zero disparity' depth that can be varied when presented whereas the
light field image has a focus depth that can be varied when presented.  For some stereoscopic 3D technologies the 'zero disparity'
depth may well be the depth of sharpest focus, for example as often seen in lenticular 3D images.

Here is an example 3D viewer that could benefit from stereoscopic information in the schema:
http://www.schillr.com/schillr3d_firefox.php5  If this software could identify the left and right images within the scope of an
images on a web page then it could reliably present them in a range of formats.

There is also a need to specific the image format for stereoscopic 3D videos.  Youtube now has some support for stereoscopic 3D
content and some search engines even allow viewing within the search results so it could be very useful to define this information.

There appears be some overlap with the need to define a schema for sub-image within a larger image.  It is common to combine a
number of images into one large image and then present different parts of this larger image across a web page - this is very common
for icons as it can reduce load time.

Some possible paths:

* Define the MediaObject 'hight and 'width to be the presented images size, not the source image object size.  Then add fields for
the top corner of the image to be presented, and perhaps have these default to center the presented image.  This would allow a
schema for the left or right images in a larger stereoscopic image and support a range of formats, such as left-right, right-left,
over-under, under-over.

* Add a stereoscopic image sequence index to the MediaObject schema.  Define this as a left to right scan, with lower indexes to the
left and higher to the right.  This is needed to order the images left to right.

* Define the 'encoding field of a MediaObject of a stereoscopic image to be a sequence of images, left to right.  It is not clear
that this is the best path.  There may be multiple encodings for a 3D images, high and low definition, left-right, and anaglyph,
etc, and overloading the 'encoding field with both the left to right image sequence and the multiple encodings may not work.

* Define a new set of schemas for images with stereoscopic images being a subset, and the light field images another.  There are
many references to images within the current schema and it would be great to be able to specify these are 3D, for example personal
photos, product photos, etc, so replace all references to an image' with an image schema.

Feedback on suitable paths would be welcomed.

Regards
Douglas Crosher


On 05/22/2012 09:10 AM, Thad Guidry wrote:
> Hmm, I am wondering how Lytro picture gallery for their "click anywhere and refocus" technology might be able to use this for
> Stereoscopic enhancement.  They have also talked about a need for this case as well.  Perhaps you can reach out to them also to find
> out what they have already found out ?   https://www.lytro.com/living-pictures/111771  
> 
> On Mon, May 21, 2012 at 9:08 AM, Douglas Crosher <dtc-schema@scieneer.com <mailto:dtc-schema@scieneer.com>> wrote:
> 
>     Has anyone considered a schema for stereoscopic 3D images?  Having a pair of images tagged with left and right could be a big help
>     to a viewer.  Could this fit within the MediaObject?  More generally, it would be useful to be able to tag a sequence of images as a
>     left/right scan fit for 3D stereoscopic viewing.  These can be presented as a wiggle, or two images selected as a single left/right
>     pair.
> 
>     There are also stereoscopic formats that fit within a single images.  For example: left/right pair, right/left pair, over/under
>     pair, checker-board, interlaced, anaglyph, etc.  Perhaps such information could be a simple extension of the MediaObject?
> 
>     When there is a single image with sub-images of a stereoscopic left-right scan then perhaps the bounding boxes of the sub-images
>     could be store, and this would handle left/right, right/left, over/under, etc.  This would also allow the information for centring
>     the sub-images to be stored.
> 
>     Regards
>     Douglas Crosher
> 
> 
> 
> 
> 
> -- 
> -Thad
> http://www.freebase.com/view/en/thad_guidry

Received on Tuesday, 22 May 2012 02:41:27 UTC