A recent article that appeared in Microscopy
Today
showed a flawed and very awkward procedure for using Adobe
Photoshop® to combine stereo pair images into a color anaglyph that can
be viewed through colored glasses. The correct procedure is quite simple
and straightforward, as described below. The specific commands and functions
shown correspond to all recent versions of Photoshop, on either a Macintosh
or Windows PC.

The starting point is, of course, a pair of stereo images.
Whether these are acquired with an electron microscope, a macro camera or by
aerial photography, the basic requirements are that they be taken at the same
magnification, be the same size, and that they include as much common area of
the specimen as possible. If the two images are taken by shifting the camera or
specimen laterally, which is the method often used at low magnifications or in
aerial photos, the horizontal shift should not be more than about 15% of the
image width. If tilting the specimen is used, angles in the 5 to 10 degree
range produce good results. Using too large a shift or tilt may still produce
images that are suitable for measurement of vertical heights, but the anaglyph
image will usually have too much displacement of features (parallax) for
comfortable viewing. In either case, the images must be viewed so that the
shifts produced by tilt or rotation are horizontal, corresponding to a vertical
orientation for the tilt axis or a horizontal shift for the displacement
method.

As shown in Figure 1, stereo viewing produces information
about the relative depth of features in the image when the viewer’s eyes
rotate inward in their sockets to focus on an object. It is only by rapidly
shifting the point of focus among many locations in a scene that we build
up a general mental map of the relative distances of objects by stereopsis.
In most scenes it is other clues – relative size or position, rate of
motion in the visual field, shading, the obscuring of one object by another,
etc. – that provide most of the information on distance, which is why
people with one eye, or who for various reasons cannot fuse stereo images,
can still function well, drive cars, even play sports. When different
images are provided to each eye, vision must select the same object in
both images. If the points are displaced vertically, it becomes very difficult
for the eyes to find them in the two different images. Also, large differences
in contrast or brightness should be avoided.



Figure 1. Exaggerated diagram of stereo viewing. An angle of
about 5-10 degrees corresponds to typical reading distances.

To assemble the stereo images into an anaglyph, open
both of them in Photoshop. The example shown in Figure 2 is a stereo pair
of a small fossil (an ammonite) using a macro camera, stopped down to
get enough depth of field to keep the surface in focus. The fossil was
mounted on modeling clay and tilted to get the required two views. The
object was shifted to bring a recognizable location near the center to
approximately the center of the field of view.


Figure
2.
Stereo pair images of an ammonite fossil.

All colored stereo viewing glasses use the convention
that the left eye has a red filter while the right eye uses blue or green.
In case you didn’t save the cheap cardboard glasses they handed out when
your local theater showed its most recent stereo thrill-a-rama (the enthusiasm
for stereo movies seems to revive about every 10 years), you can purchase
them easily. SPI Supplies sells good quality ones with real frames (Figure
3) for about $10, and both they and many other suppliers (don’t forget
eBay) will sell you a dozen of the cardboard kind for that amount.



Figure 3. The SPI Supplies plastic-framed stereo viewing glasses.

If the original images are in color, rather than
gray scale, they must first be converted. This method of stereo viewing
presents a monochrome image to each eye. (To view colored stereo images,
other methods are used such as projectors and glasses with polarizing
filters, or LCD lenses that can flicker on and off at a rate that corresponds
to the screen display, or an optical viewer that directs different images
to each eye.) Some people with experience in viewing stereo can just stare
at the side-by-side images and achieve fusion. This article is directed
toward the anaglyph method, which practically speaking restricts the images
to monochrome. In Photoshop, selecting Image->Mode->Grayscale
will remove any color information from the image and leave just the brightness
or luminance information.

Depending on the nature of the image, you may want to make
some adjustments so that the gray scale images have good contrast and
brightness. This can usually be done satisfactorily using the Image->Adjustments->Levels
function, by setting the black and white limits and the medium gray slider on
the histogram (Figure 4). If more complex adjustments are needed, the Image->Adjustments->Curves
function provides additional control. Remember that the two images should end
up with similar brightness and contrast for the easiest viewing of the
resulting anaglyph composite. This is also the time to apply any image
processing, such as sharpening the detail contrast with an unsharp mask (Filter->Sharpen->Unsharp
Mask
), if desired.



Figure 4. Setting the sliders to mark the black, white and medium
gray points on the histogram improves image contrast.

It does not matter whether the images are 8 bit or
16 bit. Recent versions of Photoshop (CS and CS2, also known as Photoshop
8 and 9) can perform copy/paste operations, adjustments, and channel operations
on either. But for viewing or printing anaglyph images there is no advantage
to the greater bit depth after any adjustments to the images have been
made, and the resulting file will be smaller if 8 bits is used. You can
convert a 16 bit image to 8 bits by selecting Image->Mode->8
bits/channel
.

Having carried out all of the preliminaries, the assembly
process is quick and straightforward. Make sure that the left eye image is the
active image (in Photoshop the title bar will be colored or highlighted).
Choose Select->All (the keyboard shortcut is Command/Control-A) and Edit->Copy
(shortcut Command/Control-C). This places the image onto the clipboard. Now make
the right eye image the active one by clicking on it. This is presently a gray
scale image, but the anaglyph requires color. So convert the image to color (Image->Mode->RGB
Color
). The appearance on screen will not change, but the image now
consists of three identical red, green and blue color channels. Make sure that
the “Channels” palette is open (if not, select Window->Channels) and
click on the name of the Red channel as shown in Figure 5. That selects that
channel as the location where the image on the clipboard can be pasted (Edit->Paste,
or the keyboard shortcut Command/Control-V).



Figure 5. Select the red channel so that the clipboard contents can be pasted into it.

You have now assembled the anaglyph. To see it, click
on the eye symbol for the RGB display (Figure 6). That will show all of
the channels, with the left eye image that you pasted into the composite
in red, while the original right eye image remains in the blue and green
channels. Using both blue and green produces a cyan color for the right
eye image that works well with all stereo viewing glasses and also prints
better and brighter than just using one of the two channels.



Figure 6. Click on the “eye” icon to display all three color channels, but note that the red
channel is still selected (the name is highlighted) so that it can be shifted for alignment.

If the images are perfectly aligned, you are finished.
But in many cases, there will be some slight vertical displacement between
the two images that should be corrected, or you may want to shift one
horizontally with respect to the other so that the offsets are reduced
and the viewing is easier (this can also affect the impression of depth
in the final result). Because the Red channel is still selected from the
step above, even though all of the channels are being viewed (probably
through your stereo glasses), you can now shift the Red channel (the left
eye image) by either dragging it with the mouse or (usually more convenient)
using the arrow keys on your keyboard. Each press of the key shifts the
image by one pixel. If you depress the shift key the motion is ten pixels.

If it is also necessary to correct for slight rotational
differences between the images you can select Edit->Transform->Rotate
at this time. It is possible to use the mouse to grab a handle on the frame
around the image and drag it, but in most cases it is easier and more
controllable to enter numeric values into the tool bar (Figure 7). Note that in
this bar you can also enter shifts, including fractional pixel amounts, and
alter the size of the image in the horizontal and vertical directions. Changing
the horizontal scale should be done with great care, as it will directly alter
the parallax displacements and thus the stereo depth information. If the two
images are not actually at the same magnification, the best procedure is to
adjust the vertical scale of the red channel (the left eye image) to match the
features in the right, and then use the same value for the horizontal scale.



Figure 7. Tool bar for image rotation, shifting and scaling. An angle
value of half a degree positive rotation has been entered.

Once the images have been aligned, you may find that
shifting the red image has left some parts of the composite image in which
the red image no longer covers the cyan one. You can eliminate the extra
area by selecting Image->Crop. Then save the final result (with
a different filename) and you are finished. This image can be viewed on
screen, printed, etc. You may find that some adjustments to the overall
image brightness or gamma produce optimum viewing, and that these will
be different for viewing the image on screen, pasting it into a Word file
or Powerpoint presentation (which will in general not preserve the contrast
and brightness) projecting it, or printing it on paper (which also may
depend on the type of printer used).

Figure 8 shows the result in this example. As a side benefit
of the color anaglyph method, when viewing the image on screen, you can
use the Channels palette to show just the left or right eye images (click
on the name of the corresponding color channel) and to switch back and
forth rapidly between them to show the image displacement due to shift
or rotation. It is also possible to turn off either the green or blue
channel (click on the corresponding eye icon) if you want just red/green
or red/blue.



Figure 8. The composite anaglyph result from Figure 2.