STEREOJETS 

NOTE:  as of June 2002, commercial production of this type of print has been discontinued.  It is still interesting to read about how it works.

The Stereojet uses polarizable inks on special transparency film.  The inks are applied using inkjet technology, hence the name. 

Fig. 6.2a  Stereojet transparancy film.  Each side has long polymer molecules that are aligned as shown.  Special dichroic polarizing ink, laid down by an inkjet printer, is made of molecules that like to lie parallel to the polymer chains on the transparency film.  As white unpolarized light passes through the sheet, twin images with orthogonal polarization emerge.  The image must be viewed with polarizing glasses.	Fig. 6.2b  Stereojet transparency on a light table.  Inset: a pair of polarizing glasses. Photographed with a digital camera, you can see the double character of the image (esp. lower left).
Fig. 6.3   Checking out a Stereojet transparency.  A light table is used for back illumination.  Reflective Stereojets are (were) available.  Stereojets are tolerant of substantial head movements, both side to side, up and down (viewer height)  and front to back.  As with any polarization technique, head twists that change the alignment of the glasses off of vertical won't do.  It is best to be centered on the image, about one or two lateral widths back for the maximum effect.  Basically only one person can view a print at a time, but exact position is not crucial.

The cost of Stereojets (when they were available) was about $200 for a 16x20, which is about the same price as getting two high-quality chemical or digital prints made at a professional lab.  This may be prohibitive for posters by students and other under-funded researchers, but for art prints maybe one can get a gallery to front the cost.  However, in the two tests I did with continuous tone (film-like) photographs, the Stereojet does not seem to have the tonal range and high contrast (snap) that is easily obtained with high quality inkjet prints on glossy paper.  For example, compare Fig. 6.2b with the original: 

Fig. 6.4.  Raw Image, Lower Antelope Canyon Note high contrast and good saturation here relative to figure 6.2b.  On the other hand, the stereojet print has excellent stereo separation (lack of ghosts) and good resolution, which in part makes up for what it lacks.   If you had not seen a good color inkjet (non-stereo) print of the original you would perhaps have thought that the Stereojet is pretty neat.  But NOT ME.

For some subject matter Stereojets may work out better.  Being a double transparency I suspect there are limits to the saturation you can get when passing light through two active layers (four in the case of the reflective print).  There have been exhibits in NYC that were well received.  In light of the above quality comparisons, however, I have decided to pursue other avenues.

PARALLAX BARRIER

An autostereo print display is possible using an old technique called the parallax barrier.  This has been revived in autostereo LCD virtual reality displays and can work with prints as well.  The disadvantage for print exhibition is that head position is important.  Some advanced virtual reality devices have head tracking hardware and algorithms to adjust the display depending on where the viewer is.  In parallax barrier prints no such auto-adjustment is easily or cost-effectively carried out, and the print display must be designed for maximum positional tolerance.  We have played around with this medium a little, and have made some prints that are tolerant to 10 or 15% variations in head position.  For example you have to stand between 3 and 3.5 feet away from a 16 x 20 print.  Only one person can view a parallax barrier print at a time, and resolution is limited by the barrier grid spacing.  However, the contrast and color can be very good.  Here is how it works:

Fig. 6.5.  Parallax Barrier An interlaced image with vertical stripes (see below) is printed with an accurate inkjet printer onto backlit film.  Each stripe (red and blue in the cartoon) contains information from the right and left images of the stereo pair. A plexiglass spacer overlays the print, and a precise ruling of vertical lines is mounted on top. When the viewer is in the proper viewing zone, which depends on the barrier lines per inch (lpi) and the thickness of the plexiglass plate, the right eye is blocked from seeing any left images. The trick is to get all the variables just right so the viewing zone is largest.  These variables are:  grid spacing, grid opening (can be less than 50%), plexiglass thickness, image stripe widths, shadow zones between the stripes.....

The grid is made by laying out a pattern of vertical stripes in Photoshop, and getting the parallel grid printed onto transparency film using an accurate black and white printer.  The composite image is made in Photoshop using its image arithmetic functions to interlace the pair.  First the parallax grid is resized to make a mask that is slightly bigger.  Then the L and R images are laid in side by side.  An example is shown below:

Fig. 6.6a .  Interlaced Parallax Barrier Print A magnification of the interlaced print for Lower Antelope Canyon, Fig. 6.4.  This one was printed on a cheap Epson Color 777 at 50 lines per inch with no shadowing between the interlaced stripes.

Fig. 6.6b  The interlaced pair with the barrier screen on a light table.  This particular test uses a composite image printed on ordinary inkjet paper.  It already seems to have better color and contrast than the Stereojet.  However, the paper (or the printer) is not stable enough.  One needs to have straight and well registered lines all across the print in order to match up uniformly with the barrier grid.

It appears the parallax barrier print may be useful.    And the barrier technique is certainly cheap, especially if the barriers can be printed by the inkjet along with the composite images (we have not tried that yet).  The interlacing can be automated within Photoshop, which means you don't have to buy an expensive interlacing program.  This may work out for science exhibits in or outside lecture halls.  Being autostereo it will capture attention!  Computer model simulations and related imagery may not suffer from the somewhat course horizontal appearance.  Beyond limited horizontal resolution (I can't imagine inkjet printers doing much better than 200 lines per inch), the major downside is that there will inevitably be a limited comfort zone for viewing.  A person will have to be centered on the print and within some critical distance.  

We had hoped to optimize this procedure for use on home inkjet printers.   It is probable that the grid density cannot be too high, but even at 50 lines per inch, a 16 by 20 will have enough information to look pretty good.  Unfortunately tests have indicated that getting precise enough rulings and interlaced prints to produce a large (16 x 20 or above) image is difficult with inexpensive plastic materials and consumer inkjet printers.  Thus we have abandoned this research.