You are on page 1of 4

Conversion of a 2D Image Into a Phantogram by Dan Jacob 12/21/05

In a nutshell, the technique I use for converting a 2D image into a Phantogram is very similar to the technique
used by many for converting a 2D image into a standard 3D image. The difference is when creating a depth map
for a standard 3D image, the scene is sliced by vertical planes, while for a phantogram the scene (more often
than not one or several objects) is sliced by planes that are parallel to the base of the objects. In effect these
slices may be considered horizontal slices.

For a standard 3D conversion each slice of the scene is shifted either to the right or to the left an amount that’s
proportional to the depth of the slice. This shifting creates the parallax that differentiates the right eye image
from the left eye image. Similarly, for a phantogram conversion each horizontal slice is shifted an amount
proportional to its height above ground zero which is the plane the object is resting on.

The type of subject best suited for a phantogram, whether created photographically or done by conversion, is
one that has an object or several objects resting on a flat plane and viewed at an angle of anywhere between
thirty to sixty degrees off the horizontal. Therefore, when selecting a 2D image for converting to a phantogram,
it’s best to select an image containing on object resting on a more or less flat plane and viewed as described
above. The cat below is a typical 2D subject for a phantogram.
The ground plane I call plane zero meaning the plane of zero height.
Then I decide on how many slices to cut the object into; the more
slices, the better the results will be, but will also require more time.
In the case of the cat I decided on twenty slices. Below I show the
first step I take in creating the phantogram which is to enumerate each
horizontal slice.
At this point I must digress a little bit. Most often I will not use the
surface on which my subject rests because it may not be suitable or
may not be dramatic enough for my purposes. I will then separate the
subject from the original background with the intent of placing it later
on a background of my choice. As you can see, the cat has been cut
away from its original background. Now back to the slicing…I assign the same number to points around the
object that are the same height from the base. This of course is entirely estimated, and with some practice one is
then able to do this quite easily. If the numbers don’t work out on the first try, then they are juggled around until
all height correspond and fit properly. They do not have to be exact of
course – just approximate.
The next step is to draw equal height contours. This is similar to the
elevation contours that you see in a topographical map. Each line
represents a particular height above the base. In the third cat picture I
have drawn the equal height
contours connecting equal numbers
that have been previously assigned.
These lines serve as a guide for
slicing the image later on. In the
graphics editor that I use
(PhotoImpact) these lines are
drawn on an overlay above the base
image. Sometimes I won’t even
draw those guide lines and will
slice the object right after height
numbers are assigned. This can be
risky especially on complex shaped
objects.
The next step is to make each slice of the object into a separate entity or “object” as it is called in PhotoImpact.
It’s important that the edge of one slice matches the edge of its neighbor and that no areas are lost. This is
because later on, after slice shifting is done, all “missing” areas will have to be filled in.

After all slices are separate objects and can be manipulated independently I am ready to shift them. Here we get
into a great deal of intuition – how much to shift. The “experts” have all kinds of formulas that tell them what
the parallax shift should be, but I play it by ear. I decide if to shift each subsequent slice by two, three, or four
times its number in pixels. Let’s say I decide to make the 3D effect quite dramatic and use the “four” multiplier.
Then slice number 8, for example, would be shifted 8 times 4 = 32 pixels. It really doesn’t matter that much
whether all slices are shifted to the left or to the right. If it’s to the left, then the resulting image will be the right
eye image and vice versa. Slice number 13 would be shifted 13 times 4 = 52 pixels.

The picture below shows the slices shifted to the left. Yes, you guessed it – all the empty areas must be filled in
invisibly, that is, imperceptibly. If the original background were to be kept, the large void would have to filled
with the background that the right eye would have seen if the cat actually moved its head to the left. What’s
normally done in this case would be to copy a portion of the existing background and paste it into the empty
area and blend it in as best as possible. So now I have the right eye image. The original image was the left eye
image.
Now we need a common background image to place the left and right eye
images on. I usually browse the Internet for background images and have
many in storage for instant use. For this phantogram I used a red brick
pavement to accentuate the grey colors of the cat.

The left and right images coincide at ground zero, at the base, that is. The higher the elevation, the more
parallax is seen i.e. the more the images diverge. So when placing the cat on the bricks, the paws are placed
exactly on the same brick in each view. The original shape of the shadow was added to give a more realistic
presentation of the scene. If you cross your eyes, you will see the cat in 3D.
However, if you tilt the paper to 45 degrees and cross your eyes you’ll see the cat in 3D alright – but squashed!
This is because tilting the paper is equivalent to squashing the whole scene – tilt the paper to 90 degrees and the
whole scene disappears completely.

But we said before that phantograms are designed to be viewed at about 45 degrees in order to see the object
realistically sitting on its base, so we need to do something to compensate for the squashing at 45 degrees. We
need to introduce some anti-squashing which is nothing more than elongating the views vertically.

Now if you tilt the paper and cross your eyes, you will see the cat in more realistic proportions.

A perfectionist would comment that the bricks should become narrower as they recede to the rear and they
would be right. That is easily accomplished, but for us here and now this is just for fun! If it looks OK, don’t fix
it!

You might also like