Stereoscopic Reconstruction "In the Round".

Oct 16
21:00

2003

Charles Douglas Wehner

Charles Douglas Wehner

  • Share this article on Facebook
  • Share this article on Twitter
  • Share this article on Linkedin

The ... of a ... pair of images from a single ... requires some thought. You cannot go back in time and make a second image.If we assume that we have a ... image, and w

mediaimage

The reconstruction of a stereoscopic pair of images from a single "flattie" requires some thought. You cannot go back in time and make a second image.

If we assume that we have a right-eye image,Stereoscopic Reconstruction "In the Round". Articles and want to create a left-eye picture, we have to bear in mind that if that left-eye view had been created in the past, it would have "sneaked" past the object being photographed or painted and have captured slivers of background invisible from the right.

So a human figure in front of a pastoral scene would have pieces of countryside fringing its left.

As wheatstone - the INVENTOR of stereoscopic imaging - pointed out in his classic 1838 lecture (http://wehner.org/3d/first/index.htm ), the perspective difference between the eyes is NEGLIGIBLE for distant scenes.

So we would keep the scene intact for the right eye, and over-paint the human figure with some imaginary background until the pastoral scene has been reinvented - at least along the left fringe of the figure.

We would crop the left-hand edge of the left eye image and the right-hand edge of the right-eye image, so that when superimposed upon each other, such as by projection with polarized light, the left image lands upon the screen about two-and-a-quarter inches (62.8 to 64.8 millimetres) to the left of the right.

The eyes, viewing perhaps through polarized spectacles, will now be obliged to view in parallel if STEREOPSIS (the lack of double-vision) is to be attained. This tells the brain that the background is at infinity.

But what of the human figure?

Perhaps we want to position that figure at six feet distance - about two metres. This would suggest that the reconstructed right-eye view must be offset eighty, eighty-one or eighty-two printer's POINTS to the right of where it originally was. We can see at once that the specification for the landscape reconstruction is a sliver of image to the figure's left that is eighty, eighty-one or eighty-two points wide.

And the specification for the figure itself will be that each point of image-shifting will represent about an inch of backward/forward movement - about two-and-a-half centimetres.

This is not very good stereo. For much of our daily lives we talk to each other at a range of about six feet - and our eyes are quite capable of resolving the DEPTH of each others features to better than an inch at this range.

Even if we have images that are NOT on the point standard - 72 points to the inch - but sharper, such as 300 or 600 dots per inch, we have problems. How do we manipulate so many dots?

Furthermore, we would really like to find a way of SUGGESTING the presence of edges BETWEEN the dots - a system of ANTI-ALIASING.

With this in mind, I have created a small program called MoveR. It moves the dots to the RIGHT, and does so in steps of one sixteenth of a dot. Thus, a white dot against a black background is capable of becoming one sixteenth darker whilst the neighbouring dot becomes one sixteenth of peak white brighter.

So dots can BLEND from one to the next, and fool the eye into believing that it can perceive depths at six feet to within about a sixteenth of an inch.

A special image has to be constructed to define the amount to right shift. This is known as the Z-AXIS file, and it complements the XY file. With X, Y and Z fully specified the entire three dimensions have been defined.

You have to make your own Z-AXIS file. After all, stereoscopic reconstruction is an ART. A machine cannot do it because a machine cannot INVENT the missing data.

However, those who have the need to reconstruct "flatties" into 3D will find my free program useful.

You can read about it at http://wehner.org ools/mover/index.htm and download the ZIP file from http://wehner.org ools/mover.zip .

It was suggested on that web-page that there really ought to be a competition for the best stereo "resurrection" of Sir Charles Wheatstone - after all, it was HIS lecture that launched 3D upon the world. That suggestion is repeated here. My own early attempts at such reconstruction can be seen on my website - I use Wheatstone as the subject.

There is also an anaglyph (red-green) making tool. Other tools can be found at http://www.wehner.org ools/