Courtyard Wash: Dan's comments

Dan Margulis

Many people spent hours on this exercise and produced scores of different versions. Several persons remarked in their work notes that they thought it the most interesting of all the case studies. I’ve produced five more par versions, #1118-1122, on which comments are welcome.

Going through the process notes revealed a clear consensus on how to approach this, namely, try one thing after another until you blunder onto something that works. Unfortunately this means that the descriptions, though lengthy, are not particularly useful.

Few people tried to get any color out of the IR versions. As for detailing, we basically split between those who found nothing useful in the IR and used the base photo only, and those taking the IRs into LAB and, after drastically increasing contrast, used the L and sometimes even the A or B to construct a new grayscale version. Those choosing the first route generally had crisper-looking transitions, while those using LAB created a more weathered look, and sometimes a blurriness due to misaligned versions.

I personally believe that the weathered look is more appropriate, and I do not object to the blurriness. These glyphs have been exposed to the elements for hundreds of years.

However, we must remember that we are not trying to fool anybody. All viewers will understand that we are doing a digital reconstruction of something whose real appearance can only be guessed at. So they will cut us enormous latitude, accepting something as ultra-conservative as #1101 as well as something as over-the-top as #1113. Both do the job of restoring the figures.

Several things about our results were unexpected, at least to me. It seems to me that the hard part is getting decent definition of the glyphs. Of our sixteen entries I felt that only three failed to do so.

OTOH I initially felt (although I changed my mind later) that twelve of the sixteen failed on grounds of color. Since the other four weren’t always that great, in picking the par I thought I was choosing the lesser of the evils, and made a couple of inexplicably poor choices. I measured a couple of points in each image, and will post them tomorrow. Here were my initial criteria:

*Utah rock formations are basically some kind of orange or warm yellow, not especially saturated. This may remind you of the King of Beasts exercise, where I laid down some limits for what kind of orange the lion’s hide could be. Here, though, a much wider range is acceptable. In the light area I measured at the lower center of the background, basically it’s got to be B>A>0, with B somewhere between 1.5A and 8A.

*In real life, this type of petroglyph is always pretty dull. Given how old these colorants are it would be a surprise to have them significantly off neutral. So I said that, in measuring a painted area, a certain amount of color is acceptable but if A+B>25 then the artwork is too colorful.

*I would assume that if anything, the glyphs would measure as slightly yellow-orange, picking up the color of their rock substrate. Nevertheless, several people had them well to the magenta side of red. I can see that a slight amount of magenta might be helpful artistically in making the glyphs more perceptible against the background, but I said that if A>B+10 it can’t be acceptable.

Our results: six individuals couldn’t manage to come within the enormous allowable color range for the background. Of these, one’s glyphs measured as too colorful and two others had them too magenta.

Four people had a decent background color, but their glyphs were too colorful, and in one case too magenta as well.

Two people had good background color and glyphs that were properly dull, but according to my numbers they were too magenta.

Having now reviewed all the entries, I’ll stick with my position that a cool background is disqualifying. The other two tests now seem overly doctrinaire.

Probably this is because we rarely see a photo where absolutely nothing is a vivid color. If more of the landscape were showing, or possibly a flower or two, we would likely detect that some of the glyphs are too colorful. But when everything is so dull we subconsciously intensify any variation. If you want to know which of these entries are the most *accurate* for color, why, that would be #1110 and #1114. But just because they’re “accurate” doesn’t make them interesting.

Likewise the question of the overly magenta glyphs. The par version meets all of my color requirements, but it’s quite monotonous. It basically has both background and glyphs as a type of brown. It would be much improved if the glyphs were more neutral; they would stand out better. But if they can’t be more neutral, then making them more magenta may be better than leaving them as is.


Returning to the question of tonal contrast, this one reminds me of the Pagodas exercise, where the successful versions all blended to make the background lighter than the pagodas. Here, it seems sensible to at some point to use the red channel for a blend, because it has the most variation between the two.

The Dehaze filter within Camera Raw seemed quite effective in bringing out these shapes.

Some people wondered why they couldn’t find the white ovals/shields anywhere in the original or IR versions. The answer is because they aren’t there any more—they’ve been erased. The vandals were more successful in getting rid of the whites than the other colorants. Therefore, we take out the dodge tool, set it to Highlights, and paint them back in. Not enough people did this.

Keeping the upper right area from plugging was difficult but most of us managed it well, whether by masking it or by making use of the AB channels in Overlay blending.

#1101 illustrates why we shouldn’t make the wall too interesting. You may find, as I do, that it’s so boring as to be a distraction in itself, but we can’t deny that it makes the glyphs stand out.

It follows that we shouldn’t try to engineer in color variation. There isn’t enough natural variation in the glyphs to make it worthwhile, so all it would do is make the wall more interesting.

In just about every other picture of Arches National Park I’d be using the MMM script to get more color variation. Here, though, I omitted that step for the reasons just described.

I was too busy congratulating myself on my own cleverness, however, to take note of a corollary: if you don’t want people to pay attention to the wall, don’t sharpen it. Furthermore, having oversharpened it as I did, I couldn’t figure out why it looked out of place. It seems pretty obvious now: if the rock surface was rough and textures, the Fremont people would not have painted on it.

Well, I’ll get it right the next time. Unfortunately, there isn’t one. But I’ll have comments on individual versions tomorrow, and meanwhile you can comment on the five additional pars if you wish.