German camouflage colors.

[Patrick]

The color for 14D3, you missed, it is too light, it is close to 15C3. I would like to see 14D3 in the pattern, save your 13D3.

The color in that sample is 14D3 as defined by the Munsell code given in the tables in the Methuen Handbook. If it looks closer to 15C3 on the actual color plates in your copy of the Handbook, the Munsell code provided in the table is incorrect.

Let Jan take a whack at it & I'll start on the lower pattern tonight.

[Patrick]

Dan, I rechecked all of the conversions for 14D3. The table in the Handbook gives a Munsell value of 9.5P 5.5 3.5. The color shown in the examples posted 07/12 is exactly what the conversion software generates.

The engineer who wrote the Munsell conversion program is adamant that the conversions from Munsell to Lab & RGB values are precise and accurate. (He also acknowledges that the CMYK values have problems because of the simplicity of the conversion formulae). My own tests confirm his statements. I believe that the conversion program is accurate and the values stated in the Methuen-Munsell tables may, in some cases, be off. Since we've found one that's questionable, there are probably others.

That being said, when I look at 14D3 on my scans of the color plates, I admit that it is a close match to Alan Toelle's color (and my plate images are by no means perfect). We could call it 14D3 and I could derive the values corresponding to the color myself. My values would have no connection with the Munsell table value in the Handbook, but they would be correct.

[Patrick]

Item 3. Five color Day Lower Pattern, Light.

I have taken a look at what alan has for the five color lower fabric. He has,
1. "A" Oriental Blue 22D5.
2. "B" Greyish Violet 15C4
3. "C" Greyish turquoise, 24C3.
4. "D" Greyish magenta 13C4.
5. "E" Ochre 5C6.
I would like to see this pattern.

Attached. This looks like the colors for Alan's "base" example (the left-side images on his Web page). "C" obviously has a problem. Not too surprising because the Handbook Munsell values are 10B 6.6 1.9; an extreme value for the Hue that is then desaturated.

[Patrick]

Two pair comparing 14D3 and 13E2.

[Patrick]

Last night I tried to figure out why a Methuen color that looks like a good match when selected off the plate can be off by quite a bit when used in the artwork. In particular, Dan San Abbott recommended 14D3 as a strong match for one of the five colors, while I came up with 13E2. I came up with my choice by converting several of the Methuen colors to L*a*b* values using the Munsell codes provided in the Handbook, then comparing the colors generated with Alan Toelle's artwork. When I went back and looked at the scans I have of the Methuen color plates, I had to agree with Dan that 14D3 looks like a very good match for one of Toelle's colors. The question is, why does it look too light and too violet in hue when I used it to produce a fabric swatch that incorprates all five colors in a pattern of irregular polygons?

Lesson Learned #1: Color is contextual as well as metrical.

My first reaction was that the Munsell values given in the Handbook are wrong; at least the one for 14D3. I've convinced myself that this statement is incorrect. I'll be happy to grind through the numbers if anyone wants to see them, but the results of a few experiments clearly show that Methuen color 14D3 does indeed correspond to Munsell code 9.5P 5.5 3.5. So I'm going back to assuming that the Methuen-Munsell tables in the Handbook are basically correct.

The best explanation that I can provide for the difference is that Dan's choice was made out-of-context of the five-color pattern. When selecting the color from a Methuen plate in an actual hardcopy of the handbook, it's impossible to know how the color will be perceived when it's surrounded by four other equally vibrant colors arrayed in a pattern. What looks great stand-alone doesn't work at all when integrated. Likewise, if you pick Toelle's fabric swatch apart into individual color chips, they look different than when they're in the pattern. You can see on the examples that I did last night that Toelle's Greyish Magenta color becomes more like a cigarette-ash grey when it's shown as a color chip. There is a good overview of this effect in the book Real World Color Management, by Fraser, Murphy & Bunting.

I'd like to hear others comment on this, whether you think it has validity or not. I believe it does, and that it has some important implications about how we communicate color information.

Lesson Learned #2: Color values should be communicated as L*a*b* values.

In talking with the engineer who wrote the Munsell Conversion program at GretagMacbeth, I realized that we're going to have problems down the road if all we provide are RGB values. The reason is that RGB is device dependent, and the identical RGB values may (probably will) look different on two different monitors. There are several reasons for this. One of them is the internal color space model used by different graphics programs. A low-end graphics program will likely not allow the user to make changes to the internal RGB color space profile -- you'll be stuck with a hardcoded sRGB space. sRGB roughly corresponds to the color space of a monitor purchased off the discount shelf at Wal Mart. I use a custom color space called BruceRGB that is suitable for a wide variety of graphics and prints well on a CMYK printer. There are many choices of color space profiles in PhotoShop and a set of RGB values will look different in each one.

Lab color is a device independent color space. It corresponds to a purely mathematical model of human color perception. If we provide L*a*b* values, anyone with a program that supports Lab color (PhotoShop being one of them) can enter them into the color picker and get the set of RGB values that are correct for their internal color space profile and attached devices. How well the color duplicates on a printer or a monitor then depends only on how accurately their devices are profiled and color-corrected to the internal space profile. I'll still provide the RGB values from my system, but with a warning that using them will only give you a good approximation of the actual color. Lab is the Rosetta Stone for communicating the color values.

[Dan_San_Abbott]

Patrick:
I don't know exactly what you are trying to accomplish. You are not reproducing the precise colors on Alan Toelle's 5 color top and bottom patterns. I have select Alan's colors because Alan has the means to test his results precisely. The values of his 5 color upper day pattern is not that different than my Methuen color values. The bottom pattern is off, compare your result with Alan's. Whatever you are doing is not working.
Jan should be with us as soon as the U.S. Mail delivers his Methuen Book. It is not my intent to get into a resach project. I still need your mailing address. I inadvertantly erased your e-mail that had your address, when I was clearing my inbox.
Blue skies,
Dan-San

[Patrick]

Dan, email of my home address is on the way.

The link between a hardcopy color plate in the Methuen Handbook and a PC is the Munsell code tables in the back of the handbook and the GretagMacbeth Munsell Conversion program that takes these values and converts them to RGB, L*a*b* and CMYK. The process of taking your observation of a Methuen value and reproducing it on a PC should be as follows:

1. Start with a Methuen value.

2. Find the corresponding Munsell code on the Handbook table.

3. Plug the code into the Munsell Conversion program and get RGB, L*a*b* and CMYK values.

4. Ideally, using these derived values in any PC graphics program should produce something close to the color shown in the Methuen Handbook color plate.

Unfortunately this is not working and I've been trying to understand why. I created the bottom pattern samples using the four-step procedure above from the Methuen codes you provided. It is obviously quite off from Alan's colors; too dark overall and the "C" color is completely wrong. I can easily reproduce Alan's colors exactly, but the result will be completely disconnected from the Munsell values provided in the Handbook.

So, the short answer to your question "what am I trying to do?" is that I'm trying to come up with an easy, step-by-step process that will allow anyone to use whatever color data we finally post and get the color you intended on their equipment. I'm trying to do this in a way that is consistent with the data provided in the Methuen Handbook, but not having a lot of success with this approach. So that leads to a series of experiments to find out what part of the four-step procedure is going wrong and fixing it.

Later this weekend I'll post an example and color data that exactly matches Alan's colors for the lower pattern. I'll ignore the Munsell codes provided in the Handbook and just do an exact match to Alan's pattern.

[Dan_San_Abbott]

Patrick;
Plain and simple it does not work. You are NOT reproducing the CORRECT colors. On the lower 5 color pattern, the ochre is correct the rest is wrong.
At the rate it 'll take 10 years to accomplish this program. I am 82, hopefully I will live to 92, but I think that is little optimistic. GOD willing and the creek don't rise, I just might.
blue skies,
Dan-San

[Patrick]

At the rate it 'll take 10 years to accomplish this program. I am 82, hopefully I will live to 92, but I think that is little optimistic. GOD willing and the creek don't rise, I just might.

Just remember that we carry our unfinished work into the next life. If you've been good, God will give you a nice shiny workstation equipped with a program that converts Methuen codes into Sanskrit, Katakana, yodeling or anything else you may need to get the job done. But if you wind up in the bad place, you'll get a ratty copy of the Handbook with only four intact plates -- and none of them are a match for the Five Color Day Lower pattern.

Probably best that we hedge our bets and get this thing done ASAP.

Check out the attachments & let me know what you think of the match to Alan Toelle's example - Item 3.

[Patrick]

Item 2. Five Color Day Upper, Dark