A lot will depend on the nature of the image, Mark. But in general creating a stereo pair from a 2D image is not likely to be convincing. It is of course possible to select areas of the image, copy them to a new layer, then shift them left or right to create the effect of depth when viewed stereoptically. But you create the problem of having to "fill" the areas behind the shifted elements in a convincing way. Not terribly difficult if the background is simple, but this will vary depending on the image.
A more difficult issue is that image elements shifted this way will still retain a flat "cut-out" look in a stereo viewer. Creating a more rounded, realistic appearance would probably require the use of 3D software to generate meshes to "support" your visual elements.
It is also possible to "paint" a Z-depth map on a new layer, using a 256-level grayscale, then save that as a new channel and bring it into a 3D rendering application that can use Z-depth information to generate a stereo pair, (lighter=closer, darker=farther away) but that still leaves you with the problem that you don’t have visual information to "fill" behind the shifted elements, and some other visual distortion is probably inevitable.
An interesting problem though…
Mark,
The whole essence of a stereo pair is that each of the two images is taken from a different viewpoint.
Normally for ‘normal’ stereo effect the difference in viewpoints is a distance of approximately 4 inches – the standard intraocular distance. By increasing this distance, stereo perspective can be increased (usually to compensate for distant viewpoints such as aerial recognisance photography).
Without images from those two different viewpoints it really isn’t possible to create the stereo effect convincingly.
That’s basically correct, Len – though a person with an intraocular distance of four inches would have a quite large head. The normal distance is closer to three than four.
Spreading the distance between the two viewpoints will increase the stereo effect, but can easily cause hyperstereo distortion as well – the so-called "puppet theater" effect. Large objects, because of the way your brain interprets 3D information, will appear to be smaller than they are, even as the 3D effect is more pronounced.
Anyone who has seen a poorly-made 3D Imax film will have had this experience. Despite the screen being 70 feet tall, the 3D effect – if produced with a poor understanding of the correct image separation distance – will create the sense one is looking into a shoebox, rather than out a huge window.
It has been determined that maintaining a ratio of about 30:1 between the image separation and distance-to-subject will produce the most "natural" stereo effect, minimizing the puppet theater problem. In other words, if the subject of your stereo pair is 30 feet away, a 1-foot separation between your images will produce a natural-appearing 3D effect. (In actual practice though I tend to push this a little, as "natural" 3D is a little bland in its impact on viewers.)
Mark,
I agree with what Dennis said. The technique he described in his first message would be your easiest approach to get something approximately useable. You would tend to have everything on separate planes, but many 3D comic books are made like that.
If you want a true 3D effect, you will have to do considerably more work, probably more than you want to do for this project. One way toward that would be to create a 3D model of the scene in your photograph, using 3D modeling software. The model could be fairly simplistic, but still going beyond the cutouts at various distances effect. You would then use Photoshop to cut out textures from the photograph to apply in the 3D program to various objects in the model. Most 3D programs let you create stereo pairs from any vantage point of your choosing.
As an alternative to building polygonal 3D models yourself, you could have people pose a reconstruction of the scene in your original photo, take pictures of that reconstruction from several vantage points and use software that creates 3D models from such pictures. Then, once again, apply the textures from the original photo to the 3D model.
What you want to do can be done, but as has been said before, it isn’t easy.
— Burton —
