Site-wide links

Imaging FAQ

 

 

I read that Ansel Adams painted the walls of his home studio a 20 percent gray color to be the best background color to display his photos. Do you have any more specific information about this color? Is it warm or cool gray? Is there an actual paint recipe for this color?   (answer)

Can fluorescent colors be printed with cymk printing system? I have Roland printer that uses eco-solmax inks. Are there any RGB or CMYK codes that will work?   (answer)

What does a white balance actually do?   (answer)

I am trying to emulate a smaller-gamut RGB color space on an LCD display using a 3x3 matrix, and I can not seem to figure out the general-purpose way of doing this. You reference Poynton´s book in a similar FAQ but he skips over the process and just spits out the answer for some important cases.   (answer)

A one dimensional halftone cell has 50% coverage with reflection density 2.0 and 50 % coverage with reflection density 1.0. How is L* calculated for this cell?   (answer)

How could I provide (or display) the approximate color of a star given it's Kelvin temperature?   (answer)

I am interested to convert a hex html or decimal rgb value to cie 1931. How can I do this?   (answer)

What is the best color to paint behind a High Defintion tv mounted on the wall?   (answer)

We measure several prints for our printer characterizations. Two questions: (1) what is best space to average - spectral reflectance, LAB or XYZ and (2) if we wanted to do a weighted average to reduce the weight of data which tended to be "inconsistent", how would we best do that?   (answer)

Are there guidelines or rules of thumb for how colors are perceived as "dominant" in an image? Are there differences in the perception of color dominance among different colors? If so, how is this described in the CIELAB color space?   (answer)

What does LCC stand for in the color science world?   (answer)

What color is chroma key paint for green screen back drops?   (answer)

How do I match the actual fabric color in PhotoshopTM? Images are digitally shot, and the final output is four color offset printing.   (answer)

I am interested to convert a hex html or decimal RGB value to CIE 1931. How can I do this?   (answer)

What would an acceptable max display luminance setting be for a display that does not see direct sunight or much ambient light so that colors and text are easily readable?   (answer)

Using the color bar on Powerpoint 2003 I drag the brightness bar on the Device RBG mode, I get a darker color of apparently the exact hue. Is the hue is truly staying the same? Also, if I drag from R 200, to R150 to R50, is it decreasing in the amount of red equally in each interval?   (answer)

How can I convert RGB or CMYK shade to Pantone?   (answer)

Can you suggest an ideal color scale for a digital elevation map to be printed in a scientific paper? Is it possible to have a color scale that looks good in color and in black and white?   (answer)

Suppose I introduce the same amount of noise in individual channels of a YUV image of 420 format (chroma has been downsampled). Which image is the most noticeable, the one with noisy Y, U, or V?   (answer)

Can you point me to the actual formula the Photoshop Hue/Sat tool is using to change given colors in an image?Can you point me to the actual formula the Photoshop (tm) Hue/Sat tool is using to change given colors in an image?   (answer)

Do you do research on ink jet printer color consistency? We are interested in printing solid colors for visual evaluations. We have a high-end ink jet and use photo paper, but we see color differences of up to 0.5 Units ΔE* between prints using the same printer, inks, profile and paper.   (answer)

Is there a way to determine what a symmetrical circle 1% dot, 120 line screen should measure? I am looking for a published table or a equation as I need to determine this on several different screen rulings.   (answer)

Is is possible to convert R, G, B values of an image captured using a flatbed scanner to densities? I can convert R, G, B to CIELab using the ICC profile of the scanner. Is there a way to convert CIELab to Densities?   (answer)

Where could an art reproduction facility obtain an optical device that used spectral-based imaging?   (answer)

What is "camera gamma" and its relationship to NTSC/HDTV video?   (answer)

This pertains to colors being used for home theater projection screens. Neutal grays have been suggested as the best color since they should not shift the color of the reflected image. Munsell grays are reported as neutral for illuminant C. What would the RGB value be for Munsell N8 and N9 under D65? One source lists N8 as 202 202 202 (L*ab- 81.3256 0.0044 -0.0087, Yxy as 59.0619 0.31272 0.32900). Is this correct?   (answer)

Can you provide a mathematical formula to convert a RGB color value in CMYK color value?   (answer)

Do you a reference for this equation used to get color decimal value from RGB values: "color_code = (r*65536) + (g*256) + b"   (answer)

How can I mathematically find the best match for a CMYK color from an array of CMYK colors?   (answer)

I have some images taken with a digital SLR at the wrong white-balance setting (tungsten instead of daylight). Is there a fixed set of RGB correction factors that I can apply to the entire batch?   (answer)

I am studying the colour of abalone shells. How can I statistically differentiate cbetween two shells? Eg: this shell is bluer than that one. I plan to use digital photos, D65K light with the camera at the normal angle.   (answer)

What is linear tiff data. What does it record?   (answer)

Why doesn't Pointer's gamut (CR&A, 1980) fully cover all real points in Munsell system as defined by real.dat?   (answer)

I heard that HDMI for HDTV's can have billions of colors but I also heard that our eyes can only see around 10 million colors. Does this mean all that extra color doesnt matter or will it seem clearer in some way?   (answer)

sRGB has a gamma of 2.2. Why do the equations have an exponent of 2.4?   (answer)

With all things being equal, is there any reason to see an increase in black dot gain simply by changing the print sequence from CMYK to KCMY?   (answer)

What is the ideal working space for digital painting in Photoshop and painter is. I tried using Adobe RGB but my reference images look so saturated.   (answer)

Is there a guideline, perhaps application-based, for the number of patches needed to make adequate ICC profiles for printer output? Is there an upper limit on the number of patches beyond which more patches do not necessarily improve the ICC profile?   (answer)

Do HSV and HSI refer to the same model? If not, what is the difference?   (answer)

What is multispectral image?   (answer)

Which RGB values or others make color that is visible only under ultraviolet ray on the white paper?   (answer)

I have an RGB digital SLR camera. Is it possible to find relationship between the camera output and the spectral reflectance of objects?   (answer)

How can I determine color names corresponding to hex color values?   (answer)

How can I extract a specified color from an RGB image using Matlab?   (answer)

How does a color TVthere is a great website describing how TV works at . work?   (answer)

How can I know the spectral sensitivity of a common CCD camera?   (answer)

How do grey scale images differ from color (RGB) images?   (answer)

How can I know the tricolor chromaticity coordinate of a common CCD camera?   (answer)

We are trying to measure color reproduction for digital cameras using the Macbeth ColorChecker. We image the Color Checker under various illuminants. Is there a better metric to calculate the color reproduction than ΔE?   (answer)

Can you tell me please what the CIELAB reading is of matte Munsell gray to be used inside color viewing cabinets for the print industry?   (answer)

I am trying to characterize my computer display. I have a failure of additivity. That is, the sum of the individual R, G, and B XYZ values does not equal the XYZ of white. Can you help?   (answer)

Assuming a monitor calibrated to the sRGB standard, is there a combination of available adjustments (gamma/contrast and/or saturation) that can be calculated and used to provide a better 'ballpark' display of an NTSC signal on an sRGB monitor?   (answer)

Can you tell me what the cmyk formula for silver is?   (answer)

Can you provide the pantone equivalent to GE plastic FXM171R-GY2A360M?   (answer)

Can you help me create the color green that is exactly in the middle of the green color spectrum using the RGB values of my display?   (answer)

In my prepress company we have to convert fabric samples by eye in to CMYK values. This is time consuming. Is there a software hardware combo that can read a piece of fabric and give the best cmyk values in a specific profile?   (answer)

What is neutral grey colour shade and the pantone number for neutral grey in pantone formula guide solid uncoated and uncoated ?   (answer)

Could you recommend a good website for the glossary of color imaging terms?   (answer)

Do you know of any research into the quantity of a particular colour eg., yellow that is in a painting and whether that painting is one that gives pleasure?   (answer)

Can you provide the CMYK formula for ANSI Safety Green?   (answer)

Are HSL values that the Paint accessory of Windows system use recognised by CIE ?   (answer)

Is there a mathematical conversion from RGB to HTML color code?   (answer)

How specifically does a magenta-dyed silver halide crystal absorb green energy? I know it has to do with wavelength ... is magenta somehow 'dovetailing' with green light to chemically interact?   (answer)

Can you tell me the name of any software for changing a gray image into color. Is any software to convert a real time gray video to color?   (answer)

I capture video camera RGB values through a microscope for image analysis of translucent microfossils. If I repeat this using a different camera I get different RGB readings. Are there any translucent standards available that I can use to calibrate between the two setups?   (answer)

I am building a color & B/W headshot photo lab. My primary light source is 5000K fluoresent tubes with a CRI rating of approx 93. Since skin tones are critical, what color should I paint the walls?   (answer)

I have a mapping website. Is there any function or algorithm that produces a collection of discrete colors that all have some minimum level of perceptual contrast between each other?   (answer)

Are there metrics that quantify whether items in an image are discriminable or identifiable?   (answer)

Can you hep me match the PPG paint color "burnt copper" with a CMYK color so I can print material that match the paint on my motorcycle?   (answer)

Which color model is best suited for tonal changes (i.e. curves) without changing the perceived color (hue, saturation; however defined)?   (answer)

Suppose that we want to measure the difference between the color of a test object and a reference object. How would I choose between using RGB color space and CIE Lab for color difference?   (answer)

We capture colour images printed on paper through a CCD camera. How can we eliminate the scattering effect of the paper and to account for optical dot gain?   (answer)

I want to create 3D look up table for CMYK data using Matlab. Can you help?   (answer)

What size print can I make with a 8 megapixel camera? If I print in high resolution (300dpi) and low resolution (72dpi). There is any formula to do that transformation from megapixel to inch?   (answer)

What is digital colour? The history of digital colour? Digital colour challenges? Problems encountered with digital colour?   (answer)

I am an artist, and I would like to display the color of the universe. Can you send me the RGB values?   (answer)

Can you help me separate Yu'v' values into Y, r-g, and b-y opponent channels for fourier analysis?   (answer)

Is there such thing as a perfect hue angle for Cyan and Magenta, in Lch?   (answer)

The LAB values output by the GretagMacbeth Munsell conversion software are different than those I can input in Photoshop. Any solution for this?   (answer)

For color rendering/reproduction purposes, is it better to use a high-end CRT or LCD monitor for accurate color creation? What are the pros and cons of each system, and why have high-end CRT manufacturers discontinued their CRT products in favor of LCD?   (answer)

How can I measure the color temperature of a image?   (answer)

Can you give me a ballpark magnitude of color difference in ΔEcmc from monitor to monitor (crt) that one might experience?   (answer)

Which is the most appropriate measurement geometry for quantitative evaluation of metallic colors? What are the variables that impact the viewing conditions (size of sample, illuminant, surround, intensity) and is there any research that addresses this questions?   (answer)

How is color temperature of a monitor set to a particular temperature, say 6300k or 9300k? Is there any specific formula for this? How can I change RGB values of a particular temperature to anothertemperature? Is there is any concept of change in offset of R,G,B?   (answer)

In a chromatic adaptation transform, how can we measure the white reference of the incoming video signals?   (answer)

What is the difference between perceptual colour gamut mapping and categorical colour gamut mapping?   (answer)

How you can relate the color temperature to RGB values of video signals? Will the color temperature vary frame by frame or will it vary pixel by pixel of the signal? Can you give the mathematical equation by which color temperature and RGB values are linked?   (answer)

Could you recommend a source for calibration tile standards for a camera-based color measurement system?   (answer)

Where do I find a definition of the term CIEJab? I found the term CIEJab in the Microsoft Longhorn specification in their document "Gamut Boundary Description and Gamut Shell.doc"   (answer)

Digital image sensors (such as those used in digital cameras)use red, green, blue ink-based color filters to generate color. Do they therefore have a color gamut that limits the range of colors that they can detect?   (answer)

I am working on a digital conversion for raw files from RGB digital cameras. Can you assist me with a formula to convert floating point data from a digital camera into LCH space and help with transforming that space through a LUT to 4-channel printer separations?   (answer)

How can I determine the whiteness/blackness of CMYK inks?   (answer)

What is the method for selecting the best combination of CMYK inks for a cmyk ink jet printer? What is the best way to determine the CMY values in order to get as large color gamut printer as possible?   (answer)

I want to calibrate my monitor by eye. Are there color charts available for download that can help me do this? I require skin tone photo, color and gray scale patches.   (answer)

What is a color management system?   (answer)

Can you help me convert between RGB or CMYK and Munsell, CIELAB, or CIEXYZ?   (answer)

What can you tell me about the blend mode algorithms?   (answer)




I read that Ansel Adams painted the walls of his home studio a 20 percent gray color to be the best background color to display his photos. Do you have any more specific information about this color? Is it warm or cool gray? Is there an actual paint recipe for this color? (903)
I am familiar with Ansel Adams comments on this topic and he wasn't too particular about the color itself, just that it have an overall reflectance of about 20% (perceptually midway between white and black in lightness). That requirement was to best reproduce the perception of contrast in his prints and assure that the dark areas were darker than the surroundings and the light areas lighter. So there is no specific gray or recipe in his recommendation, although a fairly neutral color would probably be most pleasing overall. (Back to top)


Can fluorescent colors be printed with cymk printing system? I have Roland printer that uses eco-solmax inks. Are there any RGB or CMYK codes that will work? (901)
The only way to print fluorescent colors is to use fluorescent inks. While some CMYK inks might exhibit a bit of fluorescence, it is unlikely they are fluorescent enough to produce colors that we would typically call "fluorescent". (Back to top)


What does a white balance actually do? (899)
It normalizes the RGB responses of a camera system to account for changes in the color of illumination. Ideally a proper white balance would result in R=G=B for neutral white and gray objects regardless of the illumination color. (Back to top)


I am trying to emulate a smaller-gamut RGB color space on an LCD display using a 3x3 matrix, and I can not seem to figure out the general-purpose way of doing this. You reference Poynton´s book in a similar FAQ but he skips over the process and just spits out the answer for some important cases. (886)
You might try Berns´ "Principles of Color Technology, 3rd Ed." to get a more illuminating explanation. It is hard to explain it all in a short response, but you need to first linearize your RGB values, then convert to XYZ with an appropriate 3x3 matrix, then convert to the new (smaller gamut) RGB values with the appropriate 3x3 (this one will be different from the inverse of the first or you will make no change) and then do any correction for display nonlinearities (e.g. gamma). (Back to top)


A one dimensional halftone cell has 50% coverage with reflection density 2.0 and 50 % coverage with reflection density 1.0. How is L* calculated for this cell? (843)
You first need to convert from visual density, D, to luminance factor, Y for each cell. (Y = 10^-D). Then you would average the two Y values (since both are 50% coverage, otherwise you would do a weighted average according to the areas) to get the overall Y value, which is then plugged into the L* equation. (Back to top)


How could I provide (or display) the approximate color of a star given it's Kelvin temperature? (839)
Here's a website that describes how to approximate various color temperatures in sRGB. (Back to top)


I am interested to convert a hex html or decimal rgb value to cie 1931. How can I do this? (838)
You might try easyrgb.com. (Back to top)


What is the best color to paint behind a High Defintion tv mounted on the wall? (836)
The best thing to have back there is a little bit of light that doesn't fall on the front screen (therefore washing out the image). If the lights are off in the room, then a small light behind the TV is helpful. This lightness helps to increase the perceived contrast of the display over what you see in a completely darkened room. If the room is illuminated, a light grey is probably the best choice. Light, but not too bright like a white, and neutral to not bias the color perception of the display. (Back to top)


We measure several prints for our printer characterizations. Two questions: (1) what is best space to average - spectral reflectance, LAB or XYZ and (2) if we wanted to do a weighted average to reduce the weight of data which tended to be "inconsistent", how would we best do that? (832)
There is no single "correct" answer. If you are interested in color appearance, then it makes most sense to average in a perceptual space like CIELAB. If you are interested in the physical characteristics of the printer, then it makes more sense to average in spectral reflectance. (However it would make even more sense to average in toner amounts after deriving an accurate model of the printer!). For color applications, CIELAB is probably the best choice. To do a weighted average, you could compute color differences from the mean measurement for each color and then recompute a weighted mean using a weight that is inversely proportional to those differences. (Back to top)


Are there guidelines or rules of thumb for how colors are perceived as "dominant" in an image? Are there differences in the perception of color dominance among different colors? If so, how is this described in the CIELAB color space? (830)
I am not sure what you mean by "dominant", but it would likely be the colors that occupy most of the image area and would probably be defined by hue (as in ... there is a lot of orange in that image). I don't know of any algorithms, but I would suggest looking at a histogram of hue angle in CIELAB and finding any dominant peaks. You might also search for the "Color Reproduction Index" that was published by Pointer and Hunt which does a similar division of colors. (Back to top)


What does LCC stand for in the color science world? (823)
To my knowledge, it stands for nothing. I did a little online searching and found a few references in patents to an LCC color space as a "luminance-chrominance" space. It appears the term might be used in a generic sense to refer to "any luminance chrominance space". (Back to top)


What color is chroma key paint for green screen back drops? (814)
Green! Seriously, it can be just about any color that stands out from the foreground objects or people. A high-chroma green is most commonly used, but sometimes other hues are selected. The camera/processing system is set up in the studio to use the selected color as the region to substitute in other video content. Here is a link to one source of green and blue chroma-key paint. (Back to top)


How do I match the actual fabric color in PhotoshopTM? Images are digitally shot, and the final output is four color offset printing. (810)
This is not a simple matter, but it can be accomplished with carefully calibrated and characterized imaging systems (camera, display, printer, and software). This is the domain of color management and I would suggest you look further into that area to learn more about it. "Real World Color Management" by Fraser et al. is one book that provides a useful introduction. (Back to top)


I am interested to convert a hex html or decimal RGB value to CIE 1931. How can I do this? (808)
You might try this site easyrgb.com (Back to top)


What would an acceptable max display luminance setting be for a display that does not see direct sunight or much ambient light so that colors and text are easily readable? (806)
There is no set, agreed upon, answer. However it certainly is possible to make displays that are uncomfortably bright in dim surroundings. The useful maximum is probably somewhere between 500 and 1000 candelas per square meter. (Back to top)


Using the color bar on Powerpoint 2003 I drag the brightness bar on the Device RBG mode, I get a darker color of apparently the exact hue. Is the hue is truly staying the same? Also, if I drag from R 200, to R150 to R50, is it decreasing in the amount of red equally in each interval? (804)
No, hue is not staying the same. Display devices are not linear, so that causes one problem, and even if they were, constant ratios of linear RGB wouldn't necessarily appear the same hue since the human visual system is nonlinear (although they would be closer). The only colors for which this will be approximately true are when you have only one non-zero RGB (i.e., just red, just blue, or just green being adjusted). On the second point, equal RGB intervals are not equal intervals in either amount of light (luminance) or perceived amount (brightness). (Back to top)


How can I convert RGB or CMYK shade to Pantone? (801)
Pantone is a proprietary system, so unless you license their software (or buy software from someone who has licensed their conversions), you cannot make this conversion. In addition without accurate display or printer characterization, the RGB or CMYK values would be inaccurate. (Back to top)


Can you suggest an ideal color scale for a digital elevation map to be printed in a scientific paper? Is it possible to have a color scale that looks good in color and in black and white? (769)
I'd suggest you check out the information at colorbrewer.org. It is based on good perceptual studies and should help with your needs. What you need is a scale that is monotonic in luminance/lightness so that the information is preserved in black and white images. (Back to top)


Suppose I introduce the same amount of noise in individual channels of a YUV image of 420 format (chroma has been downsampled). Which image is the most noticeable, the one with noisy Y, U, or V? (768)
Noise in the Y dimension will be most noticeable. However, YUV is often poorly defined and the differences in noise perception are often not as large as expected since there is luminance information in the U and V channels. (Back to top)


Can you point me to the actual formula the Photoshop Hue/Sat tool is using to change given colors in an image?Can you point me to the actual formula the Photoshop (tm) Hue/Sat tool is using to change given colors in an image? (767)
Sorry, I can't. The inner workings of programs like Photoshop are usually kept proprietary and it is not always obvious at all what these controls do. (Back to top)


Do you do research on ink jet printer color consistency? We are interested in printing solid colors for visual evaluations. We have a high-end ink jet and use photo paper, but we see color differences of up to 0.5 Units ΔE* between prints using the same printer, inks, profile and paper. (762)
We do some research on related topics. Given the variability in papers, inks, and the printing process, I would be very surprised to see performance better than what you are quoting. In fact, I would suspect it is difficult to print the same color on two edges of a piece of paper within a tolerance of 0.5 ΔE*. You should be very happy with the performance you are seeing. (Back to top)


Is there a way to determine what a symmetrical circle 1% dot, 120 line screen should measure? I am looking for a published table or a equation as I need to determine this on several different screen rulings. (760)
Well. that would depend on what you assume a 100% dot is. If you assume 100% means that the dots just touch when they are diagonal from one another (adjacent dots overlap), then you can take a right triangle with two sides equal to the dot spacing, compute the hypotenuse and divide that by two to get the radius of a 100% dot (skip the divide by two if you want diameter). Then dividing that by 100 will give you the radius of a 1% dot. Of course, this ignores any sort of irregularity in the dots and dot gain. (Back to top)


Is is possible to convert R, G, B values of an image captured using a flatbed scanner to densities? I can convert R, G, B to CIELab using the ICC profile of the scanner. Is there a way to convert CIELab to Densities? (754)
Since the spectral sensitivities of your scanner most likely differ from those for a given density standard, it is very unlikely you would be able to directly compute densities from the RGB measurements. The best you could do would be to construct a lookup table that would be appropriate for a given type of input medium (much like the ICC profile you mentioned). The same goes for a CIELAB to density conversion. It would be medium dependent since the two are derived with different spectral responses. (Back to top)


Where could an art reproduction facility obtain an optical device that used spectral-based imaging? (753)
There isn't a commercial implementation as yet. Several companies are considering such a device including Sinar and Geospatial Systems. At this point, you would need to implement this on your own. See the Art-SI page for details. (Back to top)


What is "camera gamma" and its relationship to NTSC/HDTV video? (751)
Gamma is the nonlinear relationship between the light intensity in the scene and the encoded camera responses that is normally described with a power function. The exponent of that power function is called "gamma". In video, the camera signal is encoded in this nonlinear way to compensate for a similar nonlinear relationship between video signal and display intensity in traditional CRT displays. However the compensation is rarely complete since it is often desired to increase the physical contrast of the video images. Hunt's text, "The Reproduction of Colour" provides much more background on this topic and Poynton's FAQ provides many answers on the confusing use of the term "gamma'. (Back to top)


This pertains to colors being used for home theater projection screens. Neutal grays have been suggested as the best color since they should not shift the color of the reflected image. Munsell grays are reported as neutral for illuminant C. What would the RGB value be for Munsell N8 and N9 under D65? One source lists N8 as 202 202 202 (L*ab- 81.3256 0.0044 -0.0087, Yxy as 59.0619 0.31272 0.32900). Is this correct? (737)
The Munsell neutral samples are quite nonselective, meaning they will remain neutral under a wide variety of illumination levels. Your numbers seem reasonable, but entirely too precise. As a good approximation, N8 will have L*a*b* values of 80, 0, 0 and N9 will have L*a*b* values 90, 0, 0. Conversion to RGB values will depend on the particular display characteristics and setup, but your values of 202,202,202 are not unreasonable. You might want to adjust them all up or down slightly if you would like a bit lighter or darker background. (Back to top)


Can you provide a mathematical formula to convert a RGB color value in CMYK color value? (728)
Unfortunately there is no single simple conversion formula to go from RGB to CMYK. It depends on the particulars of the devices and their primaries. If you do a search on "RGB to CMYK Conversion" on the internet, you will find a number of resources explaining the procedures. Here is just one example that might get you started. (Back to top)


Do you a reference for this equation used to get color decimal value from RGB values: "color_code = (r*65536) + (g*256) + b" (697)
I don't have a reference for that equation, but looking at it what it is doing is converting 3 8-bit RGB values (0-255) into a single 24 bit number where B is encoded as the least significant 8 bits, G the middle 8-bits, and R the most significant 8 bits. If that number was then expressed in hexadecimal notation, you would get the normal hex color codes that are used on the internet. (Back to top)


How can I mathematically find the best match for a CMYK color from an array of CMYK colors? (691)
As you have discovered, this is not an easy task. What you really want to do is convert your CMYK colors into a perceptual color space like CIELAB and then use color difference equations to find the nearest selections. This would require either some for of characterization or profile of your printing system. (Back to top)


I have some images taken with a digital SLR at the wrong white-balance setting (tungsten instead of daylight). Is there a fixed set of RGB correction factors that I can apply to the entire batch? (690)
Unfortunately not. There are nonlinear steps between the in-camera white balance adjustment and the final output image. Thus it is not possible to obtain the same results after the fact with a simple RGB scaling (unless you happen to have camera raw data, which is saved prior to white balance and other processing). This is illustrated by the fact that you can't get satisfactory results by trial and error. You probably can come up with a reasonable set of RGB scalars, but the results won't be the same as images collected with the proper setting. I'd suggest finding one image of the set that has something that you want to make neutral and using that to get approximate factors to apply to the rest as a first step. (Back to top)


I am studying the colour of abalone shells. How can I statistically differentiate cbetween two shells? Eg: this shell is bluer than that one. I plan to use digital photos, D65K light with the camera at the normal angle. (676)
As long as you keep your illumination, geometry, camera settings, and processing constant, then the LAB values in Photoshop would be adequate for your purposes. As far as statistics, you can use any normal statistical test on the LAB data. I would, however, suggest using multivariate tests on all three dimensions at once. You might also consider reducing your data down to color differences between shells (or some other standard). The equations for color difference computations can be found in any good text on colorimetry such as Berns' "Principles of Color Technology, 3rd Ed.". (Back to top)


What is linear tiff data. What does it record? (672)
Generally when someone refers to linear TIFF data, they are saying that the image data (normally RGB) is encoded linearly with respect to the amount of light in the scene (be it real or synthesized). Since most displays are nonlinear, it is common for image data to be nonlinear to compensate. Thus normal images are nonlinear, but in some applications it is more useful to have linear image data. (Back to top)


Why doesn't Pointer's gamut (CR&A, 1980) fully cover all real points in Munsell system as defined by real.dat? (667)
You should look at the Munsell colors in the file "1929.dat" on style="#0000aa"> available here, to compare reasonably with the Pointer gamut. Those in "real.dat" are not necessarily "real" at all. Those are any colors within the MacAdam limits. Those limits were defined by theoretical reflectance spectra that are 100% or 0% at every wavelength. This range is far more saturated than any object colors that can actually be made. The Pointer gamut is accepted as a reasonable approximation of object colors that can actually be produced. (Back to top)


I heard that HDMI for HDTV's can have billions of colors but I also heard that our eyes can only see around 10 million colors. Does this mean all that extra color doesnt matter or will it seem clearer in some way? (666)
The number of colors you can see at any given instant is fairly limited, perhaps in the tens of thousands, but your visual system is capable of adapting to different viewing conditions to make those tens of thousands depend on what is around them. For example, in a very dark scene in a movie, you might be able to distinguish many dark colors that would all look alike in a bright scene. There are also issues about how the color information is encoded, processed, and displayed that make those seemingly extra colors helpful. The bottom line is, yes, the added capabilities do indeed improve image quality. (Back to top)


sRGB has a gamma of 2.2. Why do the equations have an exponent of 2.4? (661)
The simple answer is that the equations include both linear and power function parts and in combination they approximate a simple power function with an exponent (gamma) of 2.2. In other words, the 2.4 in the sRGB equations is not a "gamma" since it is part of a more complex function. Here is a more detailed explanation. (Back to top)


With all things being equal, is there any reason to see an increase in black dot gain simply by changing the print sequence from CMYK to KCMY? (656)
Yes! There is every reason to think that the absorption and spreading of the black ink would be very dependent on what ink (or lack thereof) has come before. It makes perfect sense that you would see more dot gain if black is printed first since the black ink might spread in the paper more than it would if there were already three ink layers down. (Back to top)


What is the ideal working space for digital painting in Photoshop and painter is. I tried using Adobe RGB but my reference images look so saturated. (655)
The "ideal" space depends on the user and application. If you are finding Adobe RGB results in images that are too saturated, you might try sRGB, which has less chromatic primaries. (Back to top)


Is there a guideline, perhaps application-based, for the number of patches needed to make adequate ICC profiles for printer output? Is there an upper limit on the number of patches beyond which more patches do not necessarily improve the ICC profile? (653)
There are no set rules. I have heard of people being successful with 5x5x5x5 sampling of a CMYK process (or 5x5x5 for CMY) as long as they are careful in selecting the samples and in how they interpolate between them. It is not unusual to measure as many as 11x11x11x11 to build a good profile. The number required really depends on the printer's behavior (e.g., linearity, stability, etc.) (Back to top)


Do HSV and HSI refer to the same model? If not, what is the difference? (643)
They are slightly different and both examples of "color spaces" derived from device-dependent RGB color specifications that bear no real relationship to accurate colorimetry or color perception. They are very gross approximations, but sometimes more useful than the simple RGB specifications. You can learn more about them in texts on computer graphics or get a quick overview here. (Back to top)


What is multispectral image? (618)
"Multispectral" is an ill-defined term in imaging. In general it refers to an image that has more than 3 channels (which are typically RGB) sampling the visible spectrum (and it could extend to other wavelengths as well). When the image is sampled a relatively fine increments through the spectrum (say 20nm or less) with a bandpass equal to the sampling increment, that can be referred to as simply a "spectral image" (just like we talk about spectral reflectance or spectral power distributions). (Back to top)


Which RGB values or others make color that is visible only under ultraviolet ray on the white paper? (606)
Making a color that is visible only under UV radiation is not a matter of RGB values, it is a matter of selecting a material that is fluorescent when illuminated with UV, but matches the background under normal visible light. In that case, the RGB values would be the same as the background, but you would only use the fluorescent material for the parts you wanted to be visible under UV. (Back to top)


I have an RGB digital SLR camera. Is it possible to find relationship between the camera output and the spectral reflectance of objects? (598)
It is possible if you have some more information about the objects. For example the spectral reflectance for many objects can be fairly accurately represented with three basis functions. If you know these three basis functions, then you can use the RGB responses of the camera to determine the necessary weightings to recover a spectral reflectance curve. It would be an approximation in most cases and you can do better with more responses. (Back to top)


How can I determine color names corresponding to hex color values? (597)
This is easy to figure out yourself (your names are the best for your purposes) since there are several tools on the web for conversion between decimal and hex RGB notations. You just need to parse the hex numbers such that the first two are R, the second two G, and the last two B and use a tool like easyrgb.com to view the colors. (Back to top)


How can I extract a specified color from an RGB image using Matlab? (594)
There are two parts to this question. First, I assume you have an RGB image file to extract the data from. This file could be output from a camera, scanner, or simply an image you found on the web somewhere. Second, you need to transform the RGB data into a color space, such as CIELAB.

To extract the RGB data from an image file, use the Matlab function imread(). Details on this function can be found on this Mathworks page. This is a smart function which looks at a file and determines which procedure to use to read in the data. It will automatically handle jpeg, tiff, gif, and many other image formats. Imread() will usually return a 3D array, M rows, N columns, 3 image planes. For example: myImg = imread('myImageFile.jpg'); The RGB coordinates of a pixel are myImg[m,n,1], myImg[m,n,2], myImg[m,n,3], respectively.

To extract the actual color of that pixel, you will need more information. If you know the image is from certain scanner or camera, you can derive a characterization for that scanner which will predict CIELAB or CIEXYZ from the RGB data. If you know the data are sRGB, there are transformations to convert the data to CIELAB. If you have no other information about the RGB image file, then you cannot recover the color accurately. RGB values are device-dependent, and without information about the device, RGB data cannot be processed any further. (Back to top)


How does a color TVthere is a great website describing how TV works at . work? (580)
There is a great website describing how TV works at How Stuff Works. (Back to top)


How can I know the spectral sensitivity of a common CCD camera? (579)
You could ask the manufacturer. They might have published specs for the camera and/or sensor. Other than that you can measure them yourself, but it takes a rather elaborate laboratory setup to do so. (Back to top)


How do grey scale images differ from color (RGB) images? (571)
Grayscale images have one dimension of information for each pixel (light-dark) while RGB images have three dimensions (channels) of information (amount of red, green, and blue). There is a wide variety of methods for deriving grayscale images, but generally they are equivalent to some weighted combination of the RGB channels of a color image. (Back to top)


How can I know the tricolor chromaticity coordinate of a common CCD camera? (548)
A CCD camera does not have chromaticity coordinates. This is a common confusion between the analysis (measurement) of color and the synthesis (display) of color. Display primaries (and all displayed colors) have chromaticity coordinates. Camera signals do not have chromaticity coordinates until they are rendered to a display. With accurate analysis through characterization of the camera and accurate synthesis through characterization of the display it is sometimes possible to accurately recreate the chromaticity coordinates of the imaged scene on the display. The intermediate representation can be almost anything that preserves the data. (Back to top)


We are trying to measure color reproduction for digital cameras using the Macbeth ColorChecker. We image the Color Checker under various illuminants. Is there a better metric to calculate the color reproduction than ΔE? (521)
The ΔE metric you are using is a typical practice and probably adequate for your needs. As far as what reference to use, that would depend on your particular application. Normally, if accurate color reproduction is the goal, then color differences from the target under each of the illuminants (rather than using D65 as a reference each time) would be more appropriate. If however, you are interested in testing how well the system color corrects and white balances to produce a daylight appearance, then the color difference to the D65 reference might be more appropriate. (Back to top)


Can you tell me please what the CIELAB reading is of matte Munsell gray to be used inside color viewing cabinets for the print industry? (511)
Viewing booths typically have a Munsell N7/ background. This translates into CIELAB values of approximately L*=70, a*=0, and b*=0. (Back to top)


I am trying to characterize my computer display. I have a failure of additivity. That is, the sum of the individual R, G, and B XYZ values does not equal the XYZ of white. Can you help? (507)
What you have run into is a lack of additivity in your display and you are correct that it might be difficult to accurately characterize. However, there might be hope. For LCDs there is typically a significant black level that is always present and an additive display model can work if this black level is corrected. What you need to do is measure the black and subtract it from each primary (and the white) measurement before checking for additivity (and then add it back in to assess the final color). What your calculation is doing is incorporating that black level 3 times when you sum R + G + B and only once when you measure white. Looking at your numbers, it would appear that the black level is quite high, but that might be possible for this display. It is worth a try. If you look at the technical report from Dec. 1996 on this page you can see a fairly simple example of this procedure. (Back to top)


Assuming a monitor calibrated to the sRGB standard, is there a combination of available adjustments (gamma/contrast and/or saturation) that can be calculated and used to provide a better 'ballpark' display of an NTSC signal on an sRGB monitor? (505)
You should be able to fairly easily derive a reasonable transformation from NTSC to sRGB as you suggest. Essentially, you would use the following steps.
  1. Convert NTSC to linear RGB using the NTSC gamma definition
  2. Convert NTSC linear RGB values to sRGB linear RGB values using a 3x3 matrix
  3. Convert those linear sRGB values to nonlinear sRGB for display using the sRGB gamma definition.
Each step could be done with various degrees of precision. Step 2 will cause some out of gamut colors and might not be terribly accurate since the NTSC signal might not have been specifically coded for the NTSC primaries anyway. If you want to eliminate that conversion of RGB values, then steps 1 and 3 can approximately concatenate into a gamma adjustment on the incoming data. For a real shortcut, you could determine the required gamma adjustment visually for typical displays and apply that generally.

Poynton's book, "Digital Video and HDTV: Algorithms and Interfaces", has all the detailed specifications if you want to get that precise. (Back to top)


Can you tell me what the cmyk formula for silver is? (502)
The short answer is "no". The appearance of silver is dependent on the three-dimensional distribution of reflected light. For a flat patch of color viewed via diffuse reflectance, silver and gray are the same. Only on 3D objects, or by tilting 2D objects, can the difference between gray and silver be observed (or measured). So there isn't a simple set of CMYK values to produce silver. Instead you would need an image of a 3D object to produce the appearance of silver on a CMYK print. If you want flat printed patches to appear silver, you need to use a metallic ink. This answer also addresses the same question from the RGB perspective. (Back to top)


Can you provide the pantone equivalent to GE plastic FXM171R-GY2A360M? (486)
I'm sorry, we can't help with that. The PANTONE system is a proprietary color naming system and we have no way to convert to it even if we knew the color of the GE plastic (another proprietary designation). (Back to top)


Can you help me create the color green that is exactly in the middle of the green color spectrum using the RGB values of my display? (484)
Unfortunately there is no simple answer to your question. First of all, all RGBs are not created equal. RGB is a device-dependent measure of color and a given set of RGB values (such as 0,255,0) will be rendered as different colors on different displays. Secondly, RGB displays cannot in general produce spectral colors. They fall outside the gamut of RGB displays. The best you could do is to choose a particular display, characterize it carefully, and then compute the most chromatic gray that has the same hue as the wavelength you are interested in (hue could be defined in a color space such as CIELAB). As an alternative, you could simply approximate this by visually selecting the green you desire. (Back to top)


In my prepress company we have to convert fabric samples by eye in to CMYK values. This is time consuming. Is there a software hardware combo that can read a piece of fabric and give the best cmyk values in a specific profile? (480)
Thanks for your inquiry. It sounds like you are on the right track. What you need to do is profile your CMYK printing process to allow you to set up a transformation for CIELAB values to CMYK. You then need to measure the fabric swatches and process those measured CIELAB values through the profile to convert to CMYK. Since the fabric and the printed images will have different surface characteristics, the profile might need some fine tuning and that will likely take some trial and error. I suspect you are doing the right thing with your current product, but running into some of the limitations of reproducing color across different types of objects. Unfortunately I can't recommend any particular product that would better solve your problem, but you might want to explore options from other manufacturers to see if their performance is better in your particular situation. (Back to top)


What is neutral grey colour shade and the pantone number for neutral grey in pantone formula guide solid uncoated and uncoated ? (477)
The PANTONE system is a proprietary color naming system with no mathematical way to compute notations for desired colorimetric coordinates. You have to refer to PANTONE samples or PANTONE software to obtain the notations you require. (Back to top)


Could you recommend a good website for the glossary of color imaging terms? (475)
Here is one that seems reasonable. (Back to top)


Do you know of any research into the quantity of a particular colour eg., yellow that is in a painting and whether that painting is one that gives pleasure? (474)
I am not aware of any research on this topic. It would require some psychophysical experimentation to quantify the amount of "pleasure" (or other variable) produced by a range of paintings (and across a large population of observers) and then some image analysis to determine if there was a correlation between the amounts of any given color and the psychophysical results. Typically some sort of histogram analysis would be performed to measure the proportions of various colors in an image. This would require the definition of your color regions for categorizing the histogram (e.g. do you divide only by hue or do you use a 3D histogram). Even if there was a correlation, you would then need to explore the question of causation. Sometimes there is a correlation between two variables, but one actually does not cause the other. (Back to top)


Can you provide the CMYK formula for ANSI Safety Green? (460)
Unfortunately, I cannot help. Even if I had the CIE specification for ANSI Safety Green, the transformation to CMYK is not unique. The CMYK values required to print a given color depend on the particular inks, the paper, the printing technology, and the algorithms used to drive the printer. The Munsell designation for ANSI Safety Green is apparently 7.5G 4/9 (found in a web search, so you might want to confirm that) and the GretagMacbeth conversion freeware, gretagmacbeth.com, can give you a conversion from Munsell to a generic CMYK that would be a reasonable approximation. That's probably the best I can recommend. (Back to top)


Are HSL values that the Paint accessory of Windows system use recognised by CIE ? (456)
HSL is transform from device-dependent RGB values to an approximation of hue, saturation, and lightness. It is also device dependent and not an accurate perceptual space. Conversions can be found in most texts on computer graphics and at easyrgb.com or many other places on the internet. HSL is not recognized by the CIE. (Back to top)


Is there a mathematical conversion from RGB to HTML color code? (455)
HTML color codes are RGB values expressed in hexadecimal notation (00-FF, which is equivalent to 0-255). For example, black (0,0,0) converts to 000000, white (255,255,255) converts to FFFFFF, and a full blue (0,0,255) converts to 0000FF. A web search will turn up plenty of conversion utilities. One example is easycalculation.com. (Back to top)


How specifically does a magenta-dyed silver halide crystal absorb green energy? I know it has to do with wavelength ... is magenta somehow 'dovetailing' with green light to chemically interact? (448)
First of all, in the fully-developed film (or print) the silver halide crystal has been removed and all that remains is the dye. The magenta dye is magenta because of its spectral absorption properties. It is the fact that it absorbs green light that makes it look magenta. The reason it absorbs green light is due to its chemical structure. Essentially an allowed and stable energy level transition in the magenta dye corresponds with the energy of wavelengths that are perceived as green and those wavelengths can be absorbed. (Back to top)


Can you tell me the name of any software for changing a gray image into color. Is any software to convert a real time gray video to color? (439)
There are a number of software packages and utilities that can "colorize" images and video. A google.com search with the keywords "colorizing" and "software" will provide you with quite a list of possibilities. From a color science perspective, this process is not uniquely or correctly defined. It is simply an artistic process to associate arbitrary colors with objects in the black and white images. You would need information beyond that available in the black and white images alone to make accurate color images. (Back to top)


I capture video camera RGB values through a microscope for image analysis of translucent microfossils. If I repeat this using a different camera I get different RGB readings. Are there any translucent standards available that I can use to calibrate between the two setups? (428)
You have come upon the fundamental different between device-dependent color (e.g. RGB) and device-independent color (e.g. CIE XYZ or CIELAB). Unfortunately there is not an easy solution to your problem. Video RGB is simply not standardized to the point of providing consistent color measurement results across various cameras. Your best bet is to stick with using a single camera (or single manufacturer and model if you need more than one) to keep your measurements as consistent as possible. Barring that, you might try to make a simple linear transformation between images from the two cameras in order to approximately equate the data. This could be derived from images you collect on samples that you think are relatively uniform and stable. However, there is no guarantee that the signals from the two cameras will be linearly related to one another. (Back to top)


I am building a color & B/W headshot photo lab. My primary light source is 5000K fluoresent tubes with a CRI rating of approx 93. Since skin tones are critical, what color should I paint the walls? (424)
Since you have taken great care to specify and install high-quality illumination, then you don't want the walls to change the effective color of the illumination. That would indicate choosing a neutral gray for the wall colors. Then you need to decide how light or dark to make them. If you want the illumination to be strictly from your light sources, then make the walls darker, or even black. If you want a little more diffuse ambient illumination on your subjects, then make the walls a lighter gray. Either way, it is probably best to stick with neutral wall paint. (Back to top)


I have a mapping website. Is there any function or algorithm that produces a collection of discrete colors that all have some minimum level of perceptual contrast between each other? (421)
I am not aware of any systematic way to produce such a list of colors. You might be getting up to the point where you are simply trying to produce too many distinguishable colors. You might start with the Munsell system and use equal steps in Value (lightness from 1-10) and then select a number of hues and also trace out those Value steps at the maximum chroma. I would suspect you could reasonably generate 100 colors that way (10 grays and then 10 each of 9 different hues). GretagMacbeth (who sells the Munsell system) provides free software for the conversion from Munsell into other coordinates that might be helpful (the page is linked in several places on our FAQ).

You might also look at Cindy Brewer's work and on her linked personal research page, that describes strategies for selecting colors for maps. I don't think her examples get up to 100 colors, but they still might be helpful. (Back to top)


Are there metrics that quantify whether items in an image are discriminable or identifiable? (418)
I am not aware of formal definitions, but there does seem to be some common practice. Discriminability refers to the observers ability to detect objects in the image and determine that they are indeed different from one another. Identifiability would imply that the observers can also identify just what the objects are. Another term sometimes used in perception is detectability, which implies that the observer can tell that something is present, but not discriminate it from other objects, or identify it. Overall performance of an imaging system would have to be defined based on the application of interest and some metric the capability for the images to assist in that task. (Back to top)


Can you hep me match the PPG paint color "burnt copper" with a CMYK color so I can print material that match the paint on my motorcycle? (408)
I'm guessing from the paint name and application that this is a metallic paint (on with some metallic flakes in it to give some nice angle-dependent effects on the bike). Such paints can't be reproduced with a simple CMYK value since their appearance changes with angle. If you picked one CMYK value then you would end up with something that would look like a fairly boring brown. Companies like PPG make their marketing material by coating color cards with actual paint samples to illustrate the angle effects. In other printed material, or on their web page, you will notice that they show the paint on three-dimensional objects (either simple shapes or finished products) to communicate the color effectively. So, the best I can offer you is to reproduce some images of the bike for your marketing material and perhaps do some visual matching for other artwork.

Even for simple colors, there is probably no conversion between specific paints and CMYK or Pantone colors. If there is, you would probably have to get that from the paint manufacturer. Sorry I can't be more helpful. (Back to top)


Which color model is best suited for tonal changes (i.e. curves) without changing the perceived color (hue, saturation; however defined)? (392)
It would be best to use the CIELAB color space (LAB in Photoshop) to do these manipulations on the L*, or lightness, channel. This space is the most likely to give you perceptually meaningful results and changes in images without changing hue and chroma. (Back to top)


Suppose that we want to measure the difference between the color of a test object and a reference object. How would I choose between using RGB color space and CIE Lab for color difference? (391)
RGB is a device color space and will vary from device to device. The values have no perceptual meaning and the meaning of differences between them will vary from device to device. CIELAB is a color space that describes color in terms of human perception independent of the device used to create the color. Color differences in CIELAB have perceptual meaning. So, if you want to know about perceived color, rather than an arbitrary color designation, you need to use CIELAB. (Back to top)


We capture colour images printed on paper through a CCD camera. How can we eliminate the scattering effect of the paper and to account for optical dot gain? (389)
Essentially you can't. The scattering and optical dot gain happen before the image is captured and there is no going back. You could try some image enhancement in a program like Adobe Photoshop, but that would be an aesthetic judgement on success since getting back to original CMYK data used to make the print would depend on far too many variables (ink, paper, separation algorithm, camera sensors, etc.) to make it feasible to automate. (Back to top)


I want to create 3D look up table for CMYK data using Matlab. Can you help? (387)
I would recommend you open the "Help" window in MATLAB and type "ICC" in the search field. It will tell you about several functions in MATLAB that provide built-in support for ICC color management. You could then implement your look-up table as an ICC profile. (Back to top)


What size print can I make with a 8 megapixel camera? If I print in high resolution (300dpi) and low resolution (72dpi). There is any formula to do that transformation from megapixel to inch? (377)
First, all 8 megapixel cameras are not created equal. You need to start by finding the actual number of pixels in each dimension of the image for your camera. I looked up the Nikon Coolpix 8700, which is an 8 megapixel camera. and found that the images are 3264 x 2448 pixels. If you multiply those two numbers together you will get the total number of pixels in the image, which is just under 8 million pixels (8 megapixels) in this case. Other 8 megapixel cameras might have slightly different numbers. Then to find the print size for a given print addressibility, you just need to divide the number of pixels in each dimension of the image by the pixels per inch on the print. So at 300 dpi, this camera would make images that are about 10.9 x 8.2 inches. At 72 dpi, it would be 45 x 34 inches, and so on. I like to use 200 dpi as a rule of thumb for print quality equivalent to typical analog photographic processes without visible artifacts on the print. This camera would allow you to print 16.3 x 12.2 inch images at 200 dpi. This is roughly equivalent to a very good quality 35mm negative. It's pretty safe to say you can make prints of 8 megapixel images that will match or exceed the image quality of those you could make with 35mm film. In most cases the image quality will be significantly better than consumer 35mm film/camera combinations (assuming you are using a high-quality printer ... preferably one that prints on photographic paper).

(Note that DPI, dots per inch, is often used to describe the number of pixels per inch, PPI, in printed output. Since, depending on the printing technology, each pixel on a printed image might be made up of many dots, it is reasonable to say that the DPI of the printer exceeds the PPI of the image. Thus, PPI might be considered more precise terminology for describing images independent of technology used to render them despite the common usage of DPI and PPI interchangeably in most situations.) That's not as simple a question as it first seems. There are several ways to determine complementary colors. Perhaps the most technically accurate definition of additive complementaries is in the CIE chromaticity diagram. You would need to have a monitor characterization to convert from RGB to CIE xy and a definition of a preferred white point in xy (the complementary is defined by the white you choose) and then you could find the complementary chromaticity. Even then, you would still have undefined variables of luminance (how bright to select for the complementary) and saturation (how far from the white) unless you decided that the complementary hues had to add together to make an exact white.

If you want to work in RGB, you can make a simple approximation by subtracting each of your RGB values from the maximum for white (typically 255). This would assume that your display is linear, which they usually aren't. If you can linearize (gamma correct) the display first, your estimates will be more meaningful. So if I had a red color with RGB of 200, 50, 30, then an approximate complementary could be 55, 205, 225 (a nice cyan color). That will at least get you in the ballpark and always keep you with a simple rule that adding the color and it's complementary sums to white (255, 255, 255). (Back to top)


What is digital colour? The history of digital colour? Digital colour challenges? Problems encountered with digital colour? (347)
"Digital color" is just a term used in the color reproduction industry to refer to processes that are based on digital systems rather than historical analog systems. For example photography, printing, and television were all invented and developed as analog systems. The current state-of-the-art in each of those technologies is now digital. During the transition time people would use the term "digital color" to refer to the newer systems and the challenges posed by them that weren't present in the analog systems.

Color is color, it doesn't matter if it is produced by a digital or analog system. As such there really are no specific differences in the challenges and problems of digital color. Especially since all new systems are now digital. (Back to top)


I am an artist, and I would like to display the color of the universe. Can you send me the RGB values? (340)
The RGB or CMYK color would depend on the display or printer. The color is very close to a daylight at 5000K (standard illumination for printing). On an sRGB display (which has a white point at 6500K daylight ... bluer than the universe color) it appears beige and the RGB values are 255, 227, 210 (in R, G, B, order). (Back to top)


Can you help me separate Yu'v' values into Y, r-g, and b-y opponent channels for fourier analysis? (325)
Given Yu'v' values you already have the most important separation done. That is the separation of luminance, Y, from the chromatic components, u'v'. Since this is a linear space, you could perform your Fourier analysis direclty on the Yu'v' dimensions. Unfortunately u' and v' do not correspond directly to red-green and yellow-blue opponent dimensions. It turns out that the perceptual unique hues are not orthogonal to one another in any linear color space, so their is no easy answer. There are a variety of other color spaces used for this type of analysis and they all have their advantages and disadvantages. YCbCr is one such space used in digital video. I would suggest you look at chapters 21-26 of Charles Poynton's "Digital Video and HDTV: Algorithms and Interfaces" book for more details on some of these color spaces and transformations. (Back to top)


Is there such thing as a perfect hue angle for Cyan and Magenta, in Lch? (324)
There is no such thing as a perfect hue angle for cyan and magenta (or yellow for that matter) in CIELAB LCh or any other color space. The optimal hues for cyan, magenta, and yellow would depend on the application and the technology. Take yellow as an example. Perceptually unique yellow (a yellow with no perception of red or green in it) is at a hue angle of just about 90 degs. in CIELAB. So that is one definition of an optimal yellow. However, in a color reproduction system the best yellow to optimize the volume and nature of the color gamut might not be perceived as unique yellow. For example, in photographic systems the yellow primary tends to be a bit to the red side (lower hue angle). (Back to top)


The LAB values output by the GretagMacbeth Munsell conversion software are different than those I can input in Photoshop. Any solution for this? (312)
The CIELAB space in Adobe Photoshop is the same space as that used everywhere else including the GretagMacbeth conversion software. All I can imagine is that you are having issues with the encoding. In CIELAB, L* ranges from 0-100, but is encoded as 0-255 in Photoshop. Thus, the Photoshop L values need to be divided by 2.55 to get ordinary L* scaling. For the a* and b* dimensions, there is a simple offset. Photoshop values of 0 to 255 are equivalent to CIELAB a*b* values of -127 to +128. This allows the positive and negative values of CIELAB a* and b* to be represented as simple 8-bit (0-255) numbers in Photoshop. Thus Photoshop a*b* values need to have 127 subtracted from them to get ordinary a*b* scaling. Perhaps it is this conversion that you are missing. (Note that this conversion might only be approximately correct depending on the color management settings in your system, Photoshop, and your software versions, OS, etc.) (Back to top)


For color rendering/reproduction purposes, is it better to use a high-end CRT or LCD monitor for accurate color creation? What are the pros and cons of each system, and why have high-end CRT manufacturers discontinued their CRT products in favor of LCD? (309)
In either case, it is important to use a high-end display. You do get what you pay for. Since about 1998 or 1999 we have exclusively used high-quality LCD monitors for our critical visual experiments on color reproduction, color appearance, and image quality. That is about the best endorsement for the technology I can give. In general, LCDs are higher luminance (brighter), higher contrast (for real images), more uniform, and sharper than CRTs. The only advantages of CRTs are that they can easily handle multiple pixel resolutions, are more uniform across changes in viewing angle, and sometimes have better temporal responses. The temporal issues are largely moot now, so as long as you can work with the native LCD resolution, they are a clear better choice for critical color work.

LCDs still suffer from angular dependencies, but those are only an issue in situations with multiple observers viewing the same display. In critical situations, the viewing angle is usually easily controlled. Also, the angular dependencies have been reduced greatly in modern LCD monitors. Desktop LCDs have significantly larger color gamuts than CRTs as well (not necessarily true for laptops, since power saving trade-offs can reduce the gamut volume).

In our experience, the colorimetric characterization of LCD monitors is both more accurate and stable. You should have no concerns about switching to high-quality LCD systems for your critical color rendering and reproduction work. (Back to top)


How can I measure the color temperature of a image? (303)
Images don't really have a color temperature. Color temperature, or more properly "Correlated Color Temperature", is a measure of the color of light sources. It is based on the idea that the spectral power distribution of blackbody radiators (and therefore their color) can be defined as a function of one variable, their temperature (using Planck's equation). That temperature is measured on the absolute scale in Kelvins (K) and generally ranges from about 2000K to 10,000K. The variation in color tends to be from reddish-yellow (2000K) through white (a large range around 6000K) to bluish (10000K). The color of other light sources are measured using correlated color temperature by assigning them the color temperature of the blackbody that is the nearest color match.

So (correlated) color temperature is a reasonable short-hand way to describe the color of typical light sources and is often used to describe the white-point color of a display. Computer displays are often setup with 6500K white points to closely simulate average daylight. Older CRT displays were often more blue (around 9300K) due to technological limitations in CRTs. It is rare for displays to be set up with white points yellower than 6500K, but sometimes 5000K is used in the publishing/printing industries to match the color of the standard light sources used for evaluating prints.

Thus, an image doesn't have a color temperature, but the light source illuminating the scene does, the light source illuminating a print does, and the white point of a display does. (Back to top)


Can you give me a ballpark magnitude of color difference in ΔEcmc from monitor to monitor (crt) that one might experience? (300)
Unfortunately I cannot give you a simple answer. Two well calibrated and characterized CRT displays might agree with each other to as little as 0.5 Δcmc units for the colors that both can display. Two "off-the-shelf" CRT displays might not agree nearly so well. It is nearly impossible to guess how large the Δcmc values might get, but the displays could be different in their RGB primary colors, their white point, their tone transfer characteristics (gamma), and the characteristics of the graphics cards. These differences could result in very large Δcmc values between the displays when the same RGB values are used to drive them. I would not be at all surprised to see average Δcmc values of 20 or larger. (Back to top)


Which is the most appropriate measurement geometry for quantitative evaluation of metallic colors? What are the variables that impact the viewing conditions (size of sample, illuminant, surround, intensity) and is there any research that addresses this questions? (299)
It is generally accepted that multi-angle spectrophotometry is both necessary and sufficient for most metallic colors. However, some very strange materials might require measurements at additional angles and therefore goniospectrophotometry. With a single-geometry measurement (either 0/45 or sphere) you do not have the information necessary to distinguish a metallic sample from a non-metallic sample of a different color.

All of the variables you mentioned are important in visual assessment. Perhaps most important for metallic samples are the illumination (color and luminance) and the viewing geometry. Special viewing booths are made for evaluating metallic materials that allow careful control of the viewing geometry.

There is research on these topics, but perhaps not too much has been published. Much of it remains proprietary since the companies believe there is still competitive advantage in keeping their techniques to themselves. I would suggest looking in the literature for research on the measurement of automotive finishes to find the most information. (Back to top)


How is color temperature of a monitor set to a particular temperature, say 6300k or 9300k? Is there any specific formula for this? How can I change RGB values of a particular temperature to anothertemperature? Is there is any concept of change in offset of R,G,B? (286)
Monitors are set to specific white points through their characterization that relates RGB values sent to the display with displayed CIE XYZ tristimulus values. Specific white-point correlated color temperatures can be related to tristimulus values and then the display (or video card look-up tables) is adjusted such that maximum RGB values produce the desired white (e.g. desired XYZ values). It is sometimes done by changing the gain on RGB, but not the offset. More information on all these concepts can be found in Berns' text Principles of Color Technology, 3rd Ed. (Back to top)


In a chromatic adaptation transform, how can we measure the white reference of the incoming video signals? (279)
It is not possible to measure the white reference value for video signals. Video signals themselves do not specify a color until they are used to drive a display (whether a real or theoretical display). It is the white point of the display that provides the white reference for a chromatic adaptation transform. That is determined by driving the display with maximum RGB values and measuring the color with a spectroradiometer or colorimeter. It might also be available from reading the display specifications (if you trust the accuracy). (Back to top)


What is the difference between perceptual colour gamut mapping and categorical colour gamut mapping? (277)
These terms can mean different things to different people. However, in general you could think of categorical color gamut mapping as one form of perceptual color mapping. Categorical color mapping aims to make sure that colors in the reproduction do not cross color-naming category boundaries defined typically according to the 11 basic color terms of Berlin and Kay (white, gray, black, red, green, blue, yellow, purple, pink, orange, brown). Presumably, if the colors all remain in the same categories, the reproduction will approximately match the original perceptually. Perceptual color gamut mapping tries to best reproduce the appearances of the original image by minimizing changes in some color space such as CIELAB. Generally some rules are adopted to set priorities for the distributing the necessary errors (e.g., preserve hue, use a sigmoidal lightness mapping, and clip chroma). A more general perceptual gamut mapping algorithm wouldn't specifically concern itself with color-naming categories although the result of a good algorithm should also be reproduced colors that do not change categories from the original. (Back to top)


How you can relate the color temperature to RGB values of video signals? Will the color temperature vary frame by frame or will it vary pixel by pixel of the signal? Can you give the mathematical equation by which color temperature and RGB values are linked? (275)
Color temperature (or more correctly in this case Correlated Color Temperature) is often used to describe the color of the white point of a video display. For example 6500K is similar to an average daylight, while a higher color temperature would represent a more bluish white and a lower color temperature would represent a yellower white point. Since this is really a property of the display and not the video signals, it doesn't make sense to compute it on a frame-by-frame basis and there is no directly link between RGB values and color temperature since the same values can produce different color on different displays. (Back to top)


Could you recommend a source for calibration tile standards for a camera-based color measurement system? (265)
The answer really depends on how much money you want to spend. If you want the best (and very expensive) standards, then calibrated white and black standards are available from the National Institute of Standards and Technology (NIST) in the USA, the National Research Council of Canada, the National Physical Laboratory (NPL) in the UK (they provide a set of ceramic tiles known as BCRA tiles), and other national standards laboratories.

For many practical camera characterization questions, people use the GretagMacbeth ColorChecker Chart, which includes white and black samples among several others and has nominal calibration data. Another inexpensive option for black and white might be a Kodak gray scale target.

Alternatively, for black you could just use a box painted black on the inside and shoot it in a darkened room. It would be safe to assume that represents an ideal black with a reflectance of 0%, but maybe that's too dark for your application. (Back to top)


Where do I find a definition of the term CIEJab? I found the term CIEJab in the Microsoft Longhorn specification in their document "Gamut Boundary Description and Gamut Shell.doc" (263)
The simple answer is that there is no such thing as CIEJab as an official CIE designation. Microsoft has apparently taken it upon themselves to define a new term. Given the context you provided, I would assume they are referring to use of the CIECAM02 color appearance model without tying themselves down so specifically (you could get similar parameters using CIECAM97s). J is the lightness correlate in CIECAM02 and would be similar to the L* correlate in CIELAB. There are no a and b correlates in CIECAM02, but there are several sets of similar parameters depending on whether the dimension being described is colorfulness, chroma, or saturation. Chroma is the one most commonly used in imaging, so I would assume that Microsoft is referring to the rectangular chroma dimensions in CIECAM02, which are designated aC and bC (with the C as subscripts). These are described in Chapter 16 of the second Ed. of "Color Appearance Models". (Back to top)


Digital image sensors (such as those used in digital cameras)use red, green, blue ink-based color filters to generate color. Do they therefore have a color gamut that limits the range of colors that they can detect? (255)
Let's start with the short answer to your question; there is no such thing as a camera, or scanner, gamut. A gamut is defined as the range of colors that a given imaging device can display. To say that a camera had a gamut would be to imply that you could put a color in front of it that it could not possibly respond to. While it is certainly possible that two colors that are visually distinct might be mapped into the same color signals by a camera, that does not mean that the camera could not detect those colors. It just couldn't discriminate them. For example, a monochrome sensor will map all colors into a grayscale image and encode it as such. Certainly the encoding has a gamut (in this case a lightness range with no chroma information), but did the camera responded to all the colors put before it. It is the encoding that imposed the gamut. In the color world, encoding is based on some explicit or implied display. For example, sRGB is a description of a display and therefore defines a gamut (but only if the sRGB values are limited in range). If a camera encodes an image in sRGB, that doesn't mean that the range of colors the camera detected are only from within the sRGB display gamut, but it means the camera data have been transformed to best use that sRGB encoding. As long as a camera has three or more sensors that span the visual spectrum, then it will respond all the same stimuli as our visual system. Whether the camera can discriminate colors as well as the human visual system will depend on the encoding of the camera signals, quantitization, and the details of the camera responsivities. (To return to the black and white system, that camera encodes all the colors into a gray scale. They could then be displayed as any color within a given display, but many colors from the original scene would be mapped to the same values.)

Since there is no such thing as a gamut for an input device, then there is no way to compute it or calculate a figure of merit. Generally, the accuracy of color capture devices is assessed through the accuracy of the output values for known inputs in terms of color differences. Also, sensors are sometimes evaluate in terms of their ability to mimic human visual responses (and therefore be accurate) using quantities with names like colorimetric quality factor, that measure how close the camera responsivities are to linear transformations of the human color matching functions. Doing an internet search on "colorimetric quality factor" will lead you in the right direction. (Back to top)


I am working on a digital conversion for raw files from RGB digital cameras. Can you assist me with a formula to convert floating point data from a digital camera into LCH space and help with transforming that space through a LUT to 4-channel printer separations? (250)
First of all, I unfortunately can't provide you with a formula that does the required conversion. However, all is not lost. One reason is that the conversion is fairly complex and would be dependent on the camera, lighting, and materials being imaged. The good news is that the type of conversion you want is exactly what is done in normal color management systems. What you would need to do is first build a profile for your camera to go from RGB to CIELAB L*C*h. Typically profiles are generated to go to CIELAB L*a*b*, but the conversion to L*C*h is a simple one that could be done as an additional step. You would then want to build a second profile to go from L*C*h to separations for your unique ink combinations. The camera profile could probably be created with standard commercial software, but you might have to do some custom work for your output profile. Once you have the two profiles (which include your desired LUTs) it is relatively easy to link them together to process your images using commercial software like Adobe Photoshop. For more information, you might try the book "Real World Color Management" by Fraser, Murphy, and Bunting. There are also a lot of resources on the internet. If you do a Google search on "Color Management" you will find many sites including some that provide profile making services. One popular resource is www.chromix.com. (Back to top)


How can I determine the whiteness/blackness of CMYK inks? (246)
As it turns out the answer may be easier than you realized. While the exact answer will depend on the specific CMY inks being used, you can use Photoshop to estimate the relative lightness (or darkness depending on how you think about it) of the three colors for typical printing processes. The best metric to use is CIELAB L*, which is a measure of perceived lightness based on the relative luminance factor of the color (which is the CIE Y tristimulus value). Photoshop has an Lab mode that allows conversion between typical CMY values and CIELAB values (the Photoshop L corresponds with CIELAB L*). For typical CMYK inks, you might get approximate L* values of 62, 52, 95, and 12 respectively. This is on a scale where a perfect black is at 0 and the paper white would be at 100. The increments in between are designed to be perceptual such that an L* of 50 looks perceptually half-way between white and black in lightness. These numbers show that the yellow is very light, which is observed in the difficulty of reading yellow text on white paper. Magenta tends to be the darkest of the colored inks, but clearly not as dark as black with an L* of 12. (Back to top)


What is the method for selecting the best combination of CMYK inks for a cmyk ink jet printer? What is the best way to determine the CMY values in order to get as large color gamut printer as possible? (240)
Generally, this is accomplished through simulation using a mathematical model of the printing procedure. One would start with candidate spectral reflectance properties of inks (or model them as simple functions) and then manipulate those while optimizing the volume of the color gamut expressed in a space such as CIELAB. More recently, researchers have been following similar procedures to optimize the selection of greater numbers of inks for 6- or 8-ink printers. There has been a lot published on this topic at the IS&T/SID Color Imaging Conference over the years (see www.imaging.org) with one very recent example that includes numerous references to previous literature coming from our laboratory (Y. Chen, R.S. Berns, and L.A. Taplin, Extending Printing Color Gamut by Optimizing the Spectral Reflectance of Inks, IS&T/SID 12th Color Imaging Conference, Scottsdale, 163-169 (2004).). An abstract can be found at www.imaging.org/store/epub.cfm?abstrid=32139. (Back to top)


I want to calibrate my monitor by eye. Are there color charts available for download that can help me do this? I require skin tone photo, color and gray scale patches. (238)
This is a difficult question because an accurate monitor calibration/characterization requires both the display of desired values and typically their instrumental measurement to associate those values with the proper colors. A number of schemes have been devised to approximate this process visually. Adjusting your monitor with a test image requires an image that looks "good" or "correct" on an ideal monitor. The question then becomes, "what is the ideal monitor?" There is no single answer for that. The image you describe is much like the Kodak Q60 target that is used for calibrating scanners. So, if you could find a digital version of that that looks good on an ideal monitor, you could use it for visual adjustments. As an alternative, a Google search turned up this website that has some tips on visual and instrumental monitor setup and also includes links to images that might fulfill your needs. Since these images are designed to be "good," visual adjustment of your monitor for those images should at least bring you into the same general range as others in the field. (Back to top)


What is a color management system? (236)
Color management systems address problems associated with color reproduction within modern desktop environments. Since computers communicate only through the use of binary numbers, color peripheral devices must relate between numbers and colors. This unnatural association is at the root of a host of quality problems found in a color reproduction system. Unless some digital intervention takes place between the input side of the color reproduction process and the output side, many factors conspire to make it likely that some number of colors will be reproduced poorly. Color management systems provide a level of intelligent digital intervention to attempt to create a superior level of matched colors.

Color management systems consist of a collection of data structures, algorithms, protocols and methodologies. The industry de facto standard for color management has been specified by the International Color Consortium (ICC) ICC defines a source device as the input device which delivers an image and a destination device as the output device which renders an image. The color properties of these devices are encapsulated in an ICC data structure known as a device profile. In an ICC workflow, a Color Management Module (CMM) invokes appropriate algorithms for transforming an image delivered by a source device so that it will be rendered with appropriate colors on a destination device. Parameters for the transformation algorithms are derived by a CMM through reference to two profiles - one associated with the source device and one associated with the destination device. (Back to top)


Can you help me convert between RGB or CMYK and Munsell, CIELAB, or CIEXYZ? (234)
There is a freeware program for Munsell conversions (runs on Windows systems only) available from GretagMacbeth (the company that now sells Munsell products). Among other things, it converts from Munsell notations to XYZ, CIELAB, RGB, and CMYK. It can be found at www.gretagmacbeth.com. (Back to top)


What can you tell me about the blend mode algorithms? (232)
The specific algorithms in a commercial software package like Adobe Photoshop are usually proprietary. However, a little investigation with various images can usually be used to figure out what the program is doing, if not the actual algorithmic implementation. There is an open source program called GIMP that has similar functionality. There is an explanation of the blend mode algorithms in GIMP and a more generic and detailed explanation of blend modes with reference to Adobe Photoshop 6.0. (Back to top)


References

  1. G. Wyszecki and W.S. Stiles, Color Science: Concepts and Methods, Quantitative Data and Formulae 2nd Ed., Wiley, New York, 1982.
  2. R.W.G. Hunt, Measuring Colour 3rd Ed., Fountain Press, England, 1998.
  3. R.M. Boynton, Human Color Vision, Special Limited Edition, Optical Society of America, Washington D.C., 1992.
  4. M.D. Fairchild, Color Appearance Models Addison-Wesley, Reading, Massachusetts, 1998.
  5. R.S. Berns, Billmeyer and Saltzmann's Principles of Color Technology, Wiley, New York, NY, 2000. (from Wiley)
  6. H.G. Völz, Industrial Color Testing 2nd Ed., Wiley-VCH, Weinheim Germany, 2001.
  7. There are several links to GretagMacbeth's Color Conversion Freeware