Courthouse Wash: Results
I’ve posted the results of the Courthouse Wash exercise, our final case study for this year.
Reviewing: The panel at Courthouse Wash used to be the most spectacular display of pre-Columbian American rock art. In 1980, a vandal or vandals, using bleach and wire brushes, almost destroyed it. A quarter-century later, the late Bud Turner attempted to restore the pre-vandalism look photographically. We are given his raw captures, one a standard photograph, and several at various infrared wavelengths. Our job is to put these together in a way that suggests the grandeur of what was lost. Whether that means trying to make it look as it did pre-1980 is unclear.
We have 16 attempts, one of which came in six minutes before the deadline. There is one abstention, from a person claiming to have spent 100 hours on the project without getting anything worth showing. Most people also submitted a list of their steps, thanks very much. I haven’t read these, because I’d rather get a sense of who was successful and who wasn’t before investigating why.
The files don’t have people’s names on them, and were random-generator numbered from #1101 to #1116. As with past studies, we also have a “par” version, #1117. To get it, I chose five that I thought were among the best entrants, and averaged them, each one weighted 20%. This often creates a version that is superior to most if not all of its parents.
Normally I don't comment on results for two days after they're posted. Meanwhile, if you’d like to know how your own version stacked up, download the par version and compare the two directly. Do you think you got the same kind of quality? If not, I hope you’ll find further discussion useful.
That discussion will likely get very confused unless we can agree on terminology, because this image is full of vague objects that are difficult to describe in words. Therefore, I've made and attached a graphic with terms that I suggest we use.
The Folder is in the group's Photos section, 2021 Case Study: Courthouse Wash,
I also have zipped the terminology guide along with all entries and, due to size limiations, uploaded two file to our Files section,
Search for 051721_Courtyard-Wash_entries_Part-1.zip and 051721_Courtyard-Wash_entries_Part-2.zip
If you are going to study these versions I strongly encourage you to download it. Many of these entrants vary only in a minor way and it is hard to see the impact of a change without toggling back and forth between them.
I look forward to your comments, and hope you have found this series worthwhile.
Ok then, I'll open. Accurate, Historical. Plausible: pick two.
I'm 1102, and tossed accuracy out the back of the wagon right quick. I had a plugin for Image-J precisely designed to reveal weathered petroglyphs and used that to make masks (although for the most, I got similar masks out of lab when I looked for them). I could find no evidence of any of the original whitish pigment except in crevasses, so that I had to dodge it in. I do not feel I fully participated in the spirit of the exercise given the I ignored the IR shots after briefly playing with them.
For the petroglyph challenge, I first tried looking at channels in a number of the images, then tried some channel blending without good results.
Taking a step back, I looked closer at the IR images to see what subject features were strongest in each. Then I investigated channels of some of the IR images in different color spaces. To me, the detailing of the figures on the wall came out best in Lab color space, specifically the L channel. Stacking the IR images revealed that the one ending in 50 was out of alignment with the others, so I removed that from the stack (number mistyped in my summary sent to Dan). My entry is number 1101.Skip to bottom of message if you don't want to wade through the detailed processing steps.
I stacked the IR images with the one ending in 52 as the background image because it seemed to have the most detail across the rock wall. I added a curves layer to each IR image, used the auto button to increase contrast in L. The a and b channels seemed to have less useful information, so I made each layer monochrome with a BW adjustment. Here is how I used the IR layers:
Image ending in 52 is background layer.
Image ending in 45, added in multiply mode, blend-if to keep changes to mid-tone since most of the figures were mid-tone greys.
[Image ending in 46, no improvement in figure definition, so not used]
Image ending in 47, added in screen mode with black mask, mask painted white along top and right edge of image to bring out detail in the rocks.
[Image ending in 48: no improvement in figure definition, so not used]
Image ending in 49: added in multiply mode with black mask; painted with white brush (with brush in overlay mode) onto the figures to bring out details especially of those in the left and middle areas.
[Image ending in 51, no improvement in figure definition, so not used]
Above the IR layers, I added a curves adjustment layer in overlay mode with a mid-tones layer mask to concentrate adjustments mainly to the figures and using and blend-if to further protect lighter tones. I pulled the curve down to darken the figures, then put the curve layer into a group folder with a black mask. I then painted white on the group mask as needed to optimize the darker detail in wall figures without darkening the surrounding rock areas and without damaging the mid-tones mask inside the group (mask the mask echnique).
I then converted the Lab IR stack to sRGB (flattening in the process).
Using the flat original ending in 44 as a new base layer, I placed the sRGB combined IR result as a new layer in luminosity mode. From there:
Stamped layer, applied auto-tone with minimal change visible, so deleted the stamped layer.
Added hue/saturation layer to boost color saturation.
Made a stamped layer and sharpened with Topaz Sharpen AI.
Exported as jpg.
I see by comparison to other submissions that mine could have used more color. I found the challenge useful in that it made me think hard about ways of using the IR images to extract detail, quite different than a typical color correction task. Am looking forward to hearing how others produced their submissions.
Hi fellow masochists,
Unfortunately this exercise seems to have flummoxed several participants. I can’t say where the errors crept in since there isn’t any consistency. If we assume that “5002 Turner_IMG0044-default.jpg” is the base image, as far as size and cropping, then half of the entries are misaligned. And unfortunately the par version suffers since it appears to have included some of these errant entries. If you view the extreme right edge of the par you’ll notice double images due to merging unaligned versions. The versions that don’t align are 1101, 1104, 1105, 1106, 1107, 1109, 1113 and 1116. I’m guessing some errors were introduced by using at least one of the supplemental images. When I examined them I couldn’t find any that aligned perfectly with the base image so I decided to leave them out rather than try to force-fit them.
My entry is 1112.
HTH – Steve J
Disregarding my entry, the two suggested videos and all other possible comparisons, the two entries
I favour the most are 1103 and 1111, although I'm not thrilled with their color (1103 is a bit
pervasively orange, 1111's non-purples are very subdued, both look good with the par's color).
The best pre-1980 interpretation is 1102.
The main problem with the default image is that the main figures of interest are indistinct,
these three present those figures strongly and with acceptable colors. Better than in the par,
Many of the entries were hindered by overly busy contrast and/or lack of color differentiation.
I found it more difficult than usual to judge the entries by their own merits and not have my
thinking colored by whether or not content from the alternate exposures was incorporated into
An uninvolved viewer wouldn't care how these images came to be, just whether or not they were
interesting to look at.
However, as a participant I am interested to know whether or not infrared imaging can make
useful contributions to color imaging. Hopefully that will become evident during these
My own entry was guided by the CNHA video which showed an infrared image with detail not evident
in the pre-1980 photograph or the default image or googled images.
My original idea was to extract a comparable b&w image from the alternate exposures and marry it
with color from a straight-up correction of the default image.
The luminosity of the panel originated from IMG0047. The figures were darkened by an auxilliary
image based on a RAW conversion with white balance fiddled to give the figures true purple color
in order to differentiate them from the orange background substrate. The bordering Foreground Rock,
Ceiling and Shadow Area are (mostly) from the straight-up correction (using a selection mask).
The curse of extra time led to changes which didn't necessarily go to a better place (flatter
shadow, more contrast in the top end, more color), but going back seemed like regression.
So in for a penny, in for a pound: 1108.
As I mentioned, I don't feel I engaged with the spirit of the project. There's paying work now, so make hay while the sun shines. I allotted four hours to the project, and after two hours of hacking away thinking I'd find a clean path and redo, I saw no better way forward and just cleaned up what I'd done to make it presentable. I didn't look at the background info Dan supplied. Not happy about that either.
I think the interesting question here is about hyperspectral color, but I don't feel this is an especially fruitful example. Sadly.
I could wax on about separation techniques. Better if Dan does it, if he does. I still sometimes do reproduction static-shots through filters when I have a difficult neg. Digital cameras have weak filters.
My question is, did anyone take two paths and compare? I didn't see anything in the IR that I couldn't force the base shot to cough up.
1102. My paying job is to hide what I do to pictures. Out of habit, I made up the missing 'good shot' of what it looked like prior to desecration, not what a shot of it would look like now if that all hadn't happened. Ergo, by having oddball color shifts toward cyan in the shadows, I was able to hide my cluelessness about pigment type, density, and variable weathering. My guess is that there were two different preparations of the red earth colors. Also a copper influenced earth used to outline the circles, but it seemed this was added after the white, and didn't bind well to the rock everywhere.
"So, here is the final 2021 case study challenge: cancel the cancel culture by restoring
the Courthouse Wash petroglyphs to their pre-1980 glory. Or, more accurately, turning
the scene into something more understandable."
I think dangling the carrot of infrared imaging in front of our noses has been a distraction
from our goal, it certainly was for me.
Trying to extract useful details from the alternate exposures was interesting and all,
and it was certainly gratifying to get something comparable Mr. Turner's infrared image.
But I'm not sure using it led to "something more understandable", at least by my hands.
If everybody prefers versions that weren't supplemented by infrared information, then the
alternate exposures have no color imaging value, or we used it poorly.
In hindsight it might have been more "purposeful" (?) to have done the case study in two steps:
the first to tackle the infrared image preparation followed by correcting the base image
incorporating the results. I think the prospect of a daunting technical problem of very narrow
applicability detered participation in the more useful (and interesting) part of the exercise.
BTW, I'll certainly post details my infrared shenanigans if anyone is interested, but I think
it will fit better after Dan starts a technical discussion, when techniques can be compared.
But this should give it away: red subtract blue
I think this was the most challenging exercise.
My try is no. 1109
I tested all possible combinations, and found that only a few of the pictures matched – size and all. So I chose 2 pictures to work with: one for black and white contrast and the other for colour.
Ending in 45 (IR) and 44
I put the 45 (IR) into Lab, neutralised a and b and gave the L a High Pass filter. This gave a BW version I assumed would work well to emphasise the dull colours in no.44.
After that, most was done on no. 44 with several layers and blending modes, all in Lab. Put 45 on top (BW) at 20% Luminosity and had to keep tweaking the layers of 44 until satisfied.
Finally, I used the Sharpen 2018 of the PPW.
I chose not to paint colours or extra contrast into it or try to enhance the whites, as there was not much of them left, and perhaps not even originally.
It’s a shame that some people think it to be OK to destroy such things from the past. If left as they are, perhaps some day somebody can decipher it all, and we could actually learn something from it.
Trying to find the colours has been a learning process, and I look forward to more comments from you guys, and to Dan’s review and his comments.
Robert S Baldassano
Sent from Mail for Windows 10
Mine was 1107 and was a blend of 3 attempts at the problem. Like so many of you I treated the IR files as a stack. My main problem was a basic problem in that I rarely use Adobe RAW in PS and therefore spent too much time figuring how to do what I wanted as I was really unfamiliar with its settings. This was also due to the fact that I recently moved from PSCS6 Extended to PS2021. So in my first attempt I did a mean of the stack, but did not convert the IR. That was ver 1. Now I have done some IR work with a modified D200 and knew I should swap the Blue and red channels of the IR files. I tried that but it seemed that when I continued to work on the stack my swaps reverted to their original settings and I could not figure out why. For my third attempt I decided to convert to monochrome as this was an option, but this turned much of the picture blue. Finally I just blended the 3 options and that left me with Ver 4 posted, but also did not give me the results I had hoped for. So in summary I think my version leaves a lot to be desired and a number of submissions a much better than mine. If I had been able to figure out why my IR conversions did not hold in my averaged stack, I might have been more successful. When I have time in the future, I might revisit this to see what I am doing wrong. In all despite the frustrations, I am glad we had this to work on as if you are shooting any IR, what you learn here might be helpful in processing those files other than the obvious blue and red channel swaps. I can’t wait to hear what the more successful posts did to obtain their results.
Yeah… I was the guy who submitted 6 minutes before the deadline! No. 1103.
I’ve so wanted to contribute messages to the forum lately - but end of the school year responsibilities and a rapidly approaching exhibition have consumed all of my time. But I was determined to participate in this one! About 20 years ago, I was working at a studio and had to do just about anything that came thru the door - inclined fire damaged images and occasionally forensic work. So I was looking forward to this challenge - kind of brought back memories!
I started at the beginning of this challenge - just by playing. I knew that my strategy would be to build images for contrast, some for basic color and at least one with boosted color. Having made a range of tests, I found myself the night before the deadline with a lot of parts and the clock ticking.
One of my experiments was to take an IR image and Equalize all of the channels - and then use a Hammer to tame the contrast. I feel like I got some good use out of this. I also used the green channel for one - and curved it and again used a Hammer to tame the contrast. I may have had some use from a version that was Man from Mars color boosted. (I thought there was more color variation in the glyphs than there proved to be.)
Then is was the complicated matter of piecing all of those parts together. I never did feel like I got the whites to my satisfaction. (They were bright cyan in my color-boosted version - and I had to resort to Replace Color to get something out of it.) Whew! But I do love a challenge - and I would like to try this again to see if there are more elegant possibilities. I always think there are.
And just FYI: an anthropologist buddy at my university tells me that the latest thinking on the Freemont culture is that climate change with increasing droughts likely forced them to return to a more nomadic, hunter-gatherer life. Until widely spreading diseases from European contact decimated much of the indigenous populations.
Always things to learn and learn from. So I was grateful to be able to participate in these… so many thanks to all the brave souls who gave it a shot - and especially to Dan who never fails to amaze me!