Something About
Our Ability to See
The World In Color

1) A Significant Discovery
The Two Slides
Let's Duplicate the Experiment
A Bit of Analysis
An Oil-Painted Version
A Fascinating Viewer to Make
More Images
Quo Vadis?
Investigating Color Deficiency
Another Viewer Experiment
Retinex for Dichromats
3-D Shadowgraphs

(Click any image for a large view, which will
open in a new window. To continue with
the text, simply close that window)

5) An Oil-Painted Version

rw paint

For an undergraduate course in art which I took at Brown University, I thought a wonderfully offbeat final term project would be to apply what Land had discovered to the illustrative arts. Since no one had told me "this can't be done," I decided to attempt to render that familiar Retinex still-life in oils. The big trick is that I would limit myself to the same colors of the Land experiment: black and white for the medium filtered portions (taken through green), adding only red to these for the long filter portions (taken through red). Just three tubes of paint. What would THAT do?? I realized from experiments that there ought be a wide black border, to separate the limited color image from the rest of the real world's much greater color range. So the painting would be mounted within an opening cut in a large piece of black poster cardboard.
The painting itself was executed in pretty much the usual way, on a sheet of what's called "canvas board:" a cotton canvas glued down to a thick layer of cardboard. That takes longer to dry, as the air will only circulate from the front, but in this case it saved stretching and mounting the canvas before painting, something I'd done before and didn't want to be bothered with on a deadline. The paints were the usual Grumbacher tubes that my father and I frequently used when we wanted to fool with oil-painting. We had a bottle of linseed oil (that odor still brings back many happy memories!), a tin of turpentine, and some assorted brushes. Nothing fancy.
I measured one of the photos in the magazine article, laid out the original shapes in pencil on the canvas, and began to paint. I referred constantly to those two black and white images you saw above. Those guided me as to what proportion of white and red to add to the black, or vice-versa, to maintain the same proportions of dark and light as the original color separations contained. I deliberately toned down the background a little, so that the brighter portions, which weren't all that bright, would look lighter by contrast, and show off the color tones I hoped for. It might not even work -- yikes! I included a small light with the final canvas and black poster board, so that a beam of could be cast on the canvas, the rest of the room's lights extinguished. I wrote a short accompanying paper which both gave some of the fundamentals of the then available understanding of Retinex color vision, and suggested other ways to view the painting. So that was the project: something old, something new, something borrowed, and black & white and red all over...! ;^) Um. Got carried away there...
Alas, I underestimated the amount of work involved (the story of my life, you have no idea...), and discovered to my chagrin that I would not make the final deadline for the paper. I told lanky, laconic Professor Roberts a few days before, and he granted me a short extension over the weekend. Later he admitted that was very pleased with the paper and the painting, and had some surprisingly kind words to say. But he wasn't so kind when he marked it: "A+ for the paper (crossed out), final value due to lateness, C+"... grrr... I saw the point, but it was an ouch, and I still bet this was one of the most unusual term papers / projects handed to him for any undergraduate art class!
Before I'd done my painting it wasn't at all obvious if Land's initial discoveries, which were done "additively," combining images by projecting (or merging the light with half-mirrors), would work in an oil painting, a real risk (the something new). Painting is a world of "subtractive" colors, pigments blended in such a way that only what all colors in a mix reflect well gets to be seen as a light tone. If you add green and red lights, you see a vivid yellow. If you mix green and red paints, well, you know, you get a sort of muddy brown. If you mix quite a bit of black paint with any other color, the result is pretty murky. But shine a nearly extinguished or "blackened" light over the beam from a white light, and, well, the white remains white. Different worlds (the difference between logical "AND" and "OR," in fact...). So I had to be very selective as I tried my stunt, to be sure within the very small range of actual colors I'd be using, that the reflected shades matched closely to the proportions that colored lights would display. I was very tickled when I had finished it, and while still wet gave it a look-see with a small lamp in the darkened room -- eureka, it worked!
You can view the painting on your computer screen. (Here's an even larger version, for large monitors or to print out.) But it won't work out nearly so well as viewing a print of it in a darkened room with a small spot of light on it. Probably need to attach a black border of several inches all around, too, after you print it out for yourself to try. It also works better if you are at a little distance, so all the fovea of the eye can see it at a glance, and the Retinex process better kicks in. As I suggested, a bit of squinting also helps. Even better is to use the neat viewing light described next. But that came around 25 years later, as I didn't even suspect such an idea existed when I painted the original. I'm lucky that my parents are "packrats" like me, too. They had saved the painting in their basement, where I found it years later, around the time when I had made the special light box in the late 80's. The black poster board frame had deteriorated into a shambles. But the canvas board and oils were as fresh as new after a bit of dusting and cleaning. I masked the center main parts of the painting with some tape and cardboard, then sprayed matte black paint over the outer edge, which had been left blank originally, as it was behind the poster board. The spray paint trick worked well, and that's how you see it above.

(Top of the Page)
6) A Fascinating Viewer to Make

rg viewer

I was delighted when my December 1977 subscription issue of Scientific American arrived in the downstairs mail drop box which was located in the brownstone studio's outer door. For there on the cover was another article on color theory, and, yes, it was the "sequel" by Edwin Land of the pioneering report over 18 years earlier. Yeay! By now he and his team had moved onto more complex interactions, often involving all three primary colors, although the basics behind the way the earlier two-color process was by no means forgotten. The team had constructed special new equipment so that they could control the balance of red, green, and blue light falling onto two large panels, then measure any spot on either surface with a precision photometer equipped with a telescopic field of view and aiming viewfinder. It brought the element of the quantifiable to the qualitative results described earlier.
The panels were usually covered with a more or less random number of rectangular pieces of colored paper of every imaginable hue. When 50 or a hundred or more of these various sized color "swatches" were pasted up on the boards, usually making each paste-up a pretty good duplicate of the other, they resembled somewhat the paintings of Mondrian. So the team got to calling these colorful panels: "Mondrians." By turning the dials of variac transformers the proportion of the three colors of lights that fell on either panel could be controlled. Then one could compare the visual impression when, say, one bluish tone on one side was made to measure exactly the same as a greener one on the other side. You'd look at the two panels, and might smile to realize that the small swatch you saw to be "blue" on one side was in fact an exact match to the other one you saw as being "green." Other controlled experiments investigated the influence of surrounding blotches of color on a central one. Basic rules emerged, and many surprises, too.
One could "test" the idea that we need only two active receptors for color vision, by using the rods of the eye along with the long wave (red) cones. (No green or blue cones.) The team used an exceedingly dim green light and also a very extreme red one in an experiment devised by Land's collaborators, John McCann and Jeanne Benton. I thought the descriptions sounded particularly wild (could this be true?!), and had to "see" literally what this experiment was all about, constructing the box you see above. You can duplicate it yourself for just a few dollars and a few hours some evening. You'll need a small, flat cardboard box. The box a small parametric equalizer came in seemed to be a reasonable size, about six inches square on the front, and about two inches thick. I cut the two aperture on the top or front, using a razor blade and ruler, about 2 1/2 inches square each opening. Behind the openings I mounted two small flashlight bulbs in sockets I bought at Radio Shack. Double stick foam tape pieces made quick work of the mounting of them inside what was the bottom of the original box.
I also found a small battery holder for two "D" cells (long life with bulbs), and two 25 ohm rheostats (more foam tape), to adjust the brightness of each of the two bulbs. I scrounged a couple of knobs in my studio parts drawer, and some extra bits of cardboard and tape to rig an interior wall. That keeps the light on each side isolated from the opposite side. You can see most of this by studying the images above. The most expensive elements were the color filters. Here I wanted to be completely honest, so I bought some decent Kodak Wratten gel filters at the neighborhood photo store (you may have to order them). In the late 80's these were still only a few dollars each. I couldn't find a red gel that would pass light only at a suitably long wavelength that the red cones only could detect (the green cones overlap closely). You need slightly more than 650 nanometers, which is a very deep red. In the end I taped a sandwich of a deep purple Wratten filter (#35) with an orange one (#21), as this seemed, from studying the Wratten Filter charts, to provide light only above 650 nm, as desired. You'll obtain a very steep cut off curve from this sandwich. Yes, you can substitute any deep red and green filters for the viewer, but these definitely work the best:
RED = one from: 34, 34A, 35 or 36, plus one from: 15, 16, and on through 26.
GREEN = 58, 61, 74, 93 or 99.

(P.S. Note: while looking through Kodak's book on Wratten filters yesterday, I noticed that a #92 will also work quite well for the red, but that #70 is much better, and would duplicate the effect of the "sandwich" that I used. Now I remember -- Kodak reported to my local photo store that the #70 was temporarily out of stock back then, but you may have better luck now.)
Usual deep red filter
But add this purple
to this orange...
then use this green
CLICK HERE for more about the best filters for the viewer

For the other aperture I used a deep green near monochromat Wratten gel (#99, although a #93 is also good), the one filter, with it's peak transmission centered near 545 nanometers. There's little gained with red and green LED's, as their color output is very broad, nothing like our color filters (you'll still need the filters).The batteries in place, I flipped the small switch (forgot to mention that, but it's in the pix -- you wire it in series with the batteries), and both bulbs lighted. Yes! Then I cranked the two rheostats down (one in series with each bulb), and they faded out, clean and smoothly. Fine, ready to go! What we want to do is to extinguish all the lights in a room that you can close the shades to, perhaps the door also, and use only the narrow colors the box provides to examine daily objects, some color photos and other printed material. One good reference will be a color chart. Do you have one of those? If not, check out this one here:



Click on this small image to open a new window with a decent quality color chart jpeg in it, save that file, and print it out on a photo quality inkjet using decent paper, glossy works the best here. You should also try to find some nice, bold and simple color illustrations of various kinds. I find that many comic book illustrations work well, so I grabbed a few of my Carl Barks' "Scrooge McDuck Adventures," in Gladstone's newest beautiful editions (thought they'd come in handy some day! ;^). I also brought the painting I'd done back in college over to the viewer. The largest jpeg image of the painting available from this page is sharp enough that you might want to save a copy (it's right here) and print it out when you print the color chart. Again try to use a decent sized printed output, 8 x 10 is fine, on a reasonably high quality paper and settings. Or have a local service bureau make you a couple from the files. You don't have to print all the black border, and can just crop that off the file, if you prefer. Trim any white border on the prints off, too.
Okay, if everything's gathered together, turn on the viewer's bulbs, then turn off all the room lights. Bring the box over so that it can illuminate the color chart. Rotate the red brightness rheostat all the way down for a moment. Then turn the green rheostat gradually down until you no longer can see anything in front of you. Now bring it up again very slowly until you reach the minimum setting with which you can make out that there is a color chart in front of you. Fine. Let your eyes get more dark adapted, and try to find the setting which is about as low as is usable, not pitch black, but the way you usually can see in the darkened room at night. If you can detect any greenish tinge at all, it's up too high. But if your can't make out an outstretched hand in front of the box, it's too dim. Remember, we're trying to trigger only our low-light sensitive rods, which are monochrome, no cones, so no color at all, something like moonlight.
Next let's start to advance the brightness of the deep red light. Very slowly bring it up. At some point you ought to become aware suddenly that you are seeing "in color." You can bring it up a tad more than that, as the filter should keep it from leaking anything the green cones will detect. But once more, the trick is to keep the level as low as is comfortable when your eyes have adapted to the limited light available. Next, bring up the cover of the comic books, or other color illustrations. They ought look nearly normal, with visible blues and yellows and greens and browns. It all can seem very odd, indeed, as you've NEVER seen color imagery at such low light levels before -- remember in the near dark we see the world in black and white!
Now bring that print out of the painting I made. That ought look very similar to all the full color photos and prints, even though under normal lighting you can at once see that there's a limited palette available. But under this very controlled lighting we are using the most elementary two-color vision which comprises Land's Retinex color theory. I wish I'd had such a box to give my art professor back so many years ago, as this was the ideal way to demonstrate the wonder of such a painting experiment! But back then the small spotlight of white light in an otherwise darkened room was the best I could do, and it still works pretty well. When you get done looking at objects nearby with the viewer, try moving about a familiar darkened room using it as your only "flashlight." You may want to adjust the brightness on each color a bit higher for that, whatever follows the rules of minimal brightness needed to see both colors. Compare when the red is off with when it's back near the optimum setting. I find it always a surprise to experience what appears to be a normally colored world, but at such weak illumination. And many friends who've fooled around with this simple device have been just as astonished and amused.

(Top of the Page)
7) More Images


Here are three more examples of images which use only white and red light. The first is from an image Land used in the 1977 article, scanned and converted into retinex here. While compiling the earlier images from the original separations, I wondered what ordinary color snapshots might look like if converted into similar two channel pairs. So I grabbed the original photo a friend took of me with Nago in my lap, one from the same photo session that the index page photo came from, in fact, and another taken this past Christmas with a small digital camera. These had a lot of warm tones contrasted with cooler colors, and would be reasonable examples to get away from the more common still lifes with which Retinex is usually demonstrated. I yanked the blue channel from each original Photoshop file, then sized each to reasonable web page size, allowing for a generous black border.
The next step was to save two new files, one with only the red channel remaining (convert to grayscale), and the other with only the green channel present (again convert to grayscale). Now reopen the red image, convert it back to RGB. Select only the green channel and fill it with 100% black. Do the same with the blue channel. The RGB view ought now look like that one above in the merging example, an image that's all in shades of bright red down through black. Save that version with a new name. Reopen the green file, and change it also to RGB. I like to select the red channel and fill only to 15% with black, as this removes a slight bit of the red bias in the final view. It's exactly analogous to adjusting the color in that shot that Scientific American used on their cover, reducing the red somewhat (i.e. adding cyan) overall, so that neutral objects become neutral once more, and it affects nothing else much. Our eyes do the adjustment automatically for us when we view such a situation.
These were next combined with a horizontal grid image that I first had to make by hand, tediously, when doing the original images above. Big bore. I'll include one of these grids for any of you who wish to experiment yourselves, a file you should download and save (or drag to the desktop to save a copy, if your system is so setup). You can increase the width just by stretch-scaling the image. To obtain a taller version, open it into the window size you want and copy and paste it to fill the new height. Be VERY careful that you butt the images exactly together. You'll see it right away with such an even pattern. If the two sections are not properly aligned, there will appear a glitch in the overall grayish impression, usually a darker or lighter stripe that the rest of the grid. This is a horizontal grid, as I've been using. Reason for that is my Trinitron monitors use vertical stripes for the RGB colors, and although these are very, very fine, the can create what's called a "moire" pattern when presented with a similar vertical grid. So horizontal it is, not the same sort of problem.

h grid
Open and then save the grid image which appears, or just click here.

You will then open the grid image you got from the above file, select it all and paste it over the red image file, the one that looks overall red. Choose "darken only" and the grid will be superimposed. Flatten it (these are all Photoshop instructions, but most similar programs have similar commands). Open the white or slightly blue tinged white image and again paste the grid on top. Choose all and invert it (so it will be dark where the version over the red is light, and vice-versa). Choose "darken only" as before and flatten. Cool. Now copy one of them and paste it on top of the other. Choose "lighten only" and again flatten. You can then enlarge the canvas with 100% black by several inches and save. This version works great on screen on in a web page. I save space by converting to a 128 color gif with auto adaptive color selection on, but please, no dithering! Even if you're just going to print it out a black border of some kind will help. If you want to save expensive inkjet ink, just cut carefully so no white surround is left and paste the printout on a sheet of black construction paper. Congratulations, you've made your first Land Retinex two-color print!
Anyway, the two images that start this section (click on the thumbnails above, of course), will pretty well fill a smaller monitor screen, or most of a larger one. Scroll if you have to recenter the final image. Then grab the two short lengths of cardboard tubing you used above (you did try that already, didn't you?), put them nearly side by side, and hold them up to your eyes so that everything is blocked out except the given image. Since the grid is rather coarse, it will work better if you move back from the screen a little ways. You'll see two natural images with no manipulation which exhibit a notable range of colors, even though you're eyes are really only seeing two B&W views, one through red light, the other through white. Perhaps not astonishing, as you do really need three colors for deep rich colors, but damn interesting, just the same, don't you think?

(Top of the Page)
8) Quo Vadis?

The CIE Diagram
(Optimized simulation
for RGB monitors)

Just because this whole discussion could be squeezed into a not so large web page doesn't mean the topic is exhausted -- not by a factor of hundreds! If you want a fuller story of what's going on here, CLICK HERE. One can spend many an entertaining and thoughtful hour trying to extrapolate ideas and concept about our amazing human visual abilities. It's something like what alternative tuning should be for music composition, this can be for creating new images and graphics, providing new mediums to explore from. The processes are fundamental and underlying, not something painted on superficially. It's much more important to try to come up with something deeply compelling and provocative -- and "new" will follow. Sure wish so many artists were not so desperate to be "new," as to be "good" or "interesting." Sorry that I'm not more of a graphics person myself. A good artist might try to extend our horizons using such widened understandings. The intuitive curiosity that can blossom forth from true scientific discovery has often served as the fodder, the inspiration, for new works of art. The idea turns me on, anyway.
I had hoped for years to meet Edwin Land, to show him the undergraduate stunt of an oil painting, to ask his opinion on several questions I still had. But, no, I've always been too shy that way, and the occasion just never came up. I felt regretful when I read in the early 90's that he had died, after continuing his constant inquisitive adventures way past retirement from Polaroid. Land had formed the wonderfully undogmatic "Rowland Institute for Science" in Cambridge, MA. A good friend with a similar mindset, it turns out, had know him and could easily have arranged a meeting during the 80's. But I learned of this potential connection too late, although some of the stories were/are still lovely to hear. Rowland continues to investigate many of these "bits of business" that some of us just seemed doomed by our curiosities to explore. Better, they have the know how and funds to do it properly, instead of the meandering, homemade paths mere amateurs like me must follow.
There's another related, but also rather different story than what we've been speaking about here so far. It involves yet another viewing box experiment I've been fooling around with for nearly 20 years. I always wanted to find a way to duplicate, or at least nearly duplicate, the sensations of hue that a color dificient person (dichromat) would see. In the late 80's I did come up with a viewing light that seems to work well, and have described in on the next page. It's another tricky project, but well worth it. It's also gratifying to discover how well we can perceive a nearly full range of colors from a pallet that's been limited, as it indicates happily that a "colorblind" world (limited to just two colors) remains still quite colorful (more of this on the next page). While it will lead to problems at times, it's less of a mental & perceptual loss than is commonly assumed.
Eventually I'll return to other topics about color vision, and describe some other lively tests and demos you might enjoy fooling around with. I discovered a few years back that Edmund Scientific has made available some dichroic mirrors at modest cost and bought a few. These are multicoated small glass squares which reflect certain colors while allowing others to pass through. It's one more "neat toy and tool" (aren't they all?) that might help solve other questions which naturally arise in a meaty subject like this. Perhaps something reasonable will come of it, perhaps not. Meanwhile, don't just read through these descriptions, jump in and try them out yourself or with your more curious friends. Let us know if you find something interesting!

--Wendy Carlos
All original images scans from Land's articles, except where noted:
© 1959, 1977 Scientific American Inc. -- All Rights Reserved

Go on to Pt. III
(Top of the Page)

Back to the Wendy Carlos Home Page

Wendy Carlos, ColorVision2
© 2001-2008 Serendip LLC. No images, text, graphics or design
may be reproduced without permission. All Rights Reserved.