Color Management By D. B. Stovall 1 May 2014

Precision versus accuracy

Why color management? Color management used to be closed loop Print, evaluate, repeat until madness ensues With modern technology, each step in the process can be controlled so the desired image can be reproduced with little trouble IOW what appears on the monitor is close to what is in the print Current technology is precise (repeatable), color management helps tie that to relative accuracy

Color What is it? The reality Property of objects Property of light Occurs in the observer This is the light-object-observer model The reality Happens in all 3 as an event Sensation in the observer of the light from the light source as modified by the object

Light Behaves as both a particle (photon) and as an electromagnetic wave Wave behavior has frequency property Sometimes described in terms of wavelength (c/f) since frequency unit is unwieldy here (e.g. 750 THz) For visible light: Low freqs (long wavelengths) are red end of spectrum High freqs (short wavelengths) are blue end of spectrum About 700 nm for red, 400 nm for blue (nm = 10-9 m) Spectrum order for low high wavelength is ROYGBV IR is below red, UV is above violet

Color temperature Uses theoretical blackbody radiator heated to various temperatures If heated to certain temperatures will emit light with spectrum dependent on temperature alone Thermal energy is measured Uses degrees Kelvin K = °C + 273.15

White light Pure white light is equal amounts of photons at all freqs White light as we can obtain it is not pure but of several types Tungsten ~3000K Daylight (sunlight as modified by atmosphere) ~5000K Fluorescent When excitation of a gas occurs, electrons changing energy state downwards emit a photon at a particular frequency Usually a line or discontinuous spectrum LEDs are part of this family – beware!

Object behavior Absorbs or reflects at certain frequencies Modifies the light source like a filter Transmissive or reflective Certain types of material fluoresce In effect changes frequencies of the photons E.g. brighteners in papers changing UV to blue

Observer Color perception starts in the eye Cones responsible for color 3 types of cones, respond to long, medium, and short wavelengths Trichromancy Trichromatic retinal structure makes possible the 3 additive primaries

Opponency Retina color components do not work independently but in opponent pairs Light-dark Red-green Yellow-blue Zone theory of color 1st layer of retina has cones 2nd layer translates these into the 3 opponent signals Models incorporate both opponency and trichromancy B-Y R-G L-D Long Med Short

Additive primary colors Red, green, blue from long short wavelengths Black = no wavelengths White = all wavelengths All 3 added Can get any other color with some combination of these 3 R B G

Subtractive primary colors Cyan, magenta, yellow No good freq correlation since magenta is not part of color spectrum Subtracts wavelengths from otherwise white source Black = all wavelengths White = no wavelengths Can also get any color from these 3 R M Y B G C

Metamerism 2 different color samples producing the same stimuli in an observer Also the same color sample producing different stimuli in an observer Dependent on illumination and/or observer Color matching depends on the phenomena E.g. a chrome on a viewer versus an image on a monitor We can match under certain illumination conditions But under other conditions a mismatch will be apparent E.g. tungsten versus daylight Metamerism is what enables 4 color inks to represent the full spectrum Limited by gamut Can also occur between different types of observers E.g. scanners, cameras, and people

Colorimetry Applying a numeric model to color and color perception Current system created by CIE System components Illuminants like D50 or D65 Standard Observer like 2° color observer of 1931 Tristimulus response of human observer XYZ primary system Derived from Standard Observer Imaginary primaries, Y as luminance Distances are distorted xyY primary system Transform of XYZ Shows additive relationships Distances also distorted Uniform color spaces L*a*b* L* is lightness, a* is red/green opponency, b* is blue/yellow opponency Perceptually uniform L*U*V Not widely used today Usually CIEXYZ or CIELAB used in color spaces Difference calculations usually represented by ΔE

Model failures and color management Sometimes using colorimetry can get a color match in one color at the expense of other colors in the image Color constancy Perception of an object having a constant color even if the illuminant changes Devices do not have color constancy Color management can preserve the relationships between colors in an image Perceptual versus colorimetric renderings

Numerical color representation Either RGB or CMYK Numbers represent amount of colorant, not color Colorant is what is used to make a color Pigment, dye, light from a monitor phosphor, etc. Scanners (RGB) and cameras (RGB) for input Printers (CMYK) for output Monitors (RGB) for both input and output No 2 scanners or monitors will produce color in exactly the same way

Digitally encoding the RGB or CMYK Usually by even byte boundary (byte = 8 bits) structure 1 byte gives 256 levels (28) RGB (3 channels) gives 2563 theoretical colors (1.6 x 106) More bits increases fidelity and adds editing headroom E.g. 2 bytes (16 bits) gives 216 levels per channel Adding bits does not increase available dynamic range or produce more colors… These are controlled by the device itself …but decreasing the number of bits can reduce them!

Main variables of a color system Colorant color and brightness Monitor phosphors or printer inks Total range is color gamut White point color Black point density Tone curve Gamma curve in scanners, cameras, and monitors Dot gain curve in printers Sometimes a lookup table (LUT) used in place of a curve

Color models highlights RGB and CMYK are device specific models A given color triplet (x, y, z) will represent differently on different devices CIE models like CIELAB are device independent Represent perceived color All devices are limited by gamut and dynamic range Mismatches between devices require manipulation of some kind to match target device E.g. from digital camera to printer

Transfer functions f(x) g(x) h(x) Operate on input data to produce output data Change the data in a consistent, time-invariant way Color management is based on the concept of a transfer function

Color management systems Determine perceived color from RGB or CMYK inputs Attempt to keep colors consistent from device to device PCS = profile connection space Inputs Camera Scanner Whatever Outputs Printer Monitor Whatever PCS

Color management components PCS CIELAB or CIEXYZ are mandated by ICC (International Color Consortium) but PCS are not limited to these Profiles Can be for a device, class of devices, or abstract color space Basically a lookup table or mathematical transform Describes behavior but does not alter the device CMM (color management module) Software engine Converts from RGB or CMYK to PCS using data in the profile Several different ones in use ICC compliant ones are interchangeable but can differ subtly

Profile flow Data CMM Adjusted data Profile Rendering intent

Rendering intents Handles out of gamut situations E.g. camera to printer Perceptual preserves color relationships, alters all Saturation keeps colors saturated and vivid Good for graphics Relative colorimetric maps white of sources to destination and clips out of gamut colors Preserves more of the original colors than perceptual Absolute colorimetric same as relative but does not map white point Mainly for proofing

Using profiles If the image has no profile Assigning a profile is for use within that application Embedding a profile attaches to the file so the profile is available for use within different applications Assigning or embedding does not change colorant values, just how they are interpreted If an image already has an embedded profile Converting a profile for an image does change the colorant values Need to specify a target profile

Profile types Input Display Output Device space to PCS Backward transform Scanners and digital cameras Display Device space to PCS and back Forward and backward transform Monitors Output 2 way transform like display Printers and presswork

Profile internals Either 3x3 matrix or LUT Matrix LUT Uses CIEXYL For input or display LUT Also for input or display Profile size much larger Required for output profile Adds 4th channel Usually at least 6 tables Perceptual, relative colorimetric, saturation 1 for each direction

Building a profile Sending known color values to a device and see what is actually measured Monitors are generally profiled using a colorimeter Printers profiled with either a colorimeter of spectrophotometer Use known targets such as IT8 Profiles are only as accurate as measurements and only describe a gamut, not enlarge it

Families of profiles Device specific Generic Color space profiles Parameters are locally measured Generic Constructed from average device behavior or average media characteristics Not as good as specific but may be adequate Generic monitor profiles the least useful due to inherent unstable behavior Color space profiles E.g. CIELAB or CIEXYZ Device independent profiles are similar, useful for editing Adobe 98, EktaSpace, ProPhoto Typically wider gamut except for sRGB

Profiling versus calibration Characterizing a device or media Describes the device Calibration Sets the device to target characteristics Controls the device As devices change over time, must recalibrate and/or re-profile to make sure response will be as expected

Display calibration “Display” consists of monitor, video drivers, video card or HW Calibration adjusts 4 things White luminance White color Black luminance Not all calibration systems adjust this Response curve CRT monitors once in wide use, easier to calibrate due to control of electron guns Now CRTs are gone and LCD/LEDs predominate LCDs adjust with both monitor control and video LUT adjust In video control SW

Display calibration methods Visual E.g. Adobe Gamma application Pretty much useless but beats nothing Bundled monitor and calibrator E.g. LaCie Blue-Eye One button calibration May or may not include colorimeter (“puck”) Standalone calibration packages Useful on any monitor but rely on manual control of monitor and video SW Usually includes puck

Viewing environment Good idea to use a monitor hood Ambient light affects light level so use a hood and a low light level in the room Some users paint walls gray and do other things but is more important to have a consistent environment

Calibrating a monitor - 1 User inputs White point, e.g. 5000K or 6500K, or maybe a direct K input My own viewing hood was measured at 5300K so I use that Some say always use 6500K but YMMV Gamma Either 2.2 or 1.8 Most use 2.2 today 1.8 was originally used by Mac to match to LaserWriter Black point if available I use 0.2 Cd/m2 Make sure the monitor is set the way you want before you start Resolution, refresh rate, etc.

Calibrating a monitor - 2 SW will set white luminance first Some apps do this automatically, some use a user desired set point I use 120 Cd/m2 Then black luminance Can be iterative process if controlling the monitor/SW is being done manually But SW should walk you through it Then color temperature Again either set by user or automatically to target point by SW Lastly the SW displays color patches and the puck reads and feeds back measurements so profile can be built Make sure the profile is saved in a place the OS can find it Windows is /System32/Spool/Drivers/Color Mac depends on OS Once you calibrate and profile a monitor, do not do any further adjustments to the monitor or video SW or you invalidate everything!

Output (printer) profiles - 1 Must have a measuring instrument of some kind Reflective spectrophotometer is the best Best to use a device the profiling SW can talk to Try to get 4 mm to 8 mm measurement aperture Handheld ones are cheaper but measuring swatches can be tedious XY plotter types do this automatically but are expensive Do not bother with printer profilers that involve using your scanner

Output (printer) profiles - 2 General flow Read master target Print target from file Read printed target SW builds profile Verify by printing a target using new profile and reading Will probably be at least 300 swatch reads and may have to do several so SW can average Usually SW package provides target and target file If using SW that does not support measuring instrument, have to read into text file or spreadsheet and import into SW Good luck with that! Targets are IT8.7/3 or proprietary Make sure there are no profiles assigned or imbedded into target file

Output (printer) profiles - 3 When printing ensure no printer driver controls are in use In Photoshop, select Photoshop manages colors In printer driver, select no color management A profile will be for 1 printer/paper combination Changing anything requires a new profile Be aware of things like drying time Save the profile as stated above Can use canned ones from paper manufacturer as well May have to tweak your prints occasionally but will be close I use them with good results

Input (scanner) profiles Camera profiling difficult unless using controlled lighting in studio Also very brand specific Transparency and reflective only No color negative targets available Useless since orange mask varies with exposure Need physical target and target description file (TDF) Individual TDF from specific target is best but most expensive More often use TDF from target “run” Common transmissive is IT8.7/1 Usually 1 target suffices for all films except Kodachrome since dye structures are similar Common reflective is IT8.7/2 Make sure to use consistent settings and all equipment is warmed up Turn off all adjustments ICE and GEM do not generally interfere so can leave on if desired Scan the target and let the SW build the profile Save to the proper place as above I have had good luck with the canned profiles from Epson but YMMV

Evaluating what you have done Depends on viewing conditions 5000K viewing hood with adjustable intensity is ideal ICC profiles are based on D50 illuminant Some profiling applications also measure viewing light and factor that into the profile Before utilizing print to monitor matching Match brightness, not color temperature Do not put monitor and viewing hood in same field of view I violate this with good results since I set monitor white point to color temp of viewer but YMMV Various methods of validating the profiles and calibration too complex to go into in this presentation Excellently summarized in Chapter 9 of Real World Color Management by Fraser, Murphy, Bunting Book is a bit dated now but still an excellent primer on the subject Beyond this, look at workflow and specific techniques, especially for press work

Profiling monochrome images At this point does not seem to be industry agreement on the process Can present as RGB but errors in profiling may put on a slight color cast Since I do not have to deal with this I need to do more research in this area

More information www.cambridgeincolour.com/tutorials/color-management1.htm www.xrite.com/documents/literature/en/L11-176_Guide_to_CM_en.pdf www.lacie.com/us/technologies/technology.htm?id=10029 spyder.datacolor.com/