Color matching method and apparatus - Patent Review 5579031

Description Submit all comments and votes

COPYRIGHT NOTICE

A portion Of the disclosure of this patent document contains material which is subject to copyright protection. The copyright owner has no objection to the facsimile reproduction by anyone of the patent document or the patent disclosure, as it appears in the Patent and Trademark Office patent file or records, but otherwise reserves all copyright rights whatsoever.

BACKGROUND OF THE INVENTION

1. Field of the Invention

This invention relates to color matching and more particularly to a process and related apparatus for matching the color displayed on a plurality of color display devices.

2. Description of Related Art

The primary use of this invention is in the field of computer assisted color publishing systems particularly in the area of color matching. In such systems typically a color image is scanned using a scanning device which measures light intensity reflected or transmitted from the image to generate electronic signals representing three achromatic images of the original, each of the three achromatic images resulting from detecting the light intensity from the original color image through a colored filter. The filters used are almost always Red, Green and Blue. The electronic signal from the scanner is converted to a digital signal in which light intensity levels are represented as numbers. Information identifying each set of numbers representing the image information obtained through each filter is also preserved. Thus, through these steps, the original colored image is converted to a plurality of image values, and for each picture element in the picture, there are three such values, a Red, a Green and a Blue.

The image represented by the image values may be displayed in a CRT type monitor, or may be printed using a printing device able to accept electronic input. Often the image will be displayed at different locations and times using more than one display device, such as a plurality of CRT displays for observation and study by more than one operator. Similarly, hard copies may be desired in two or more different locations using two or more different printers. Even though the input values to the multiple CRT displays are the same, the displayed colors are vastly different, as anyone is well aware of who has ever observed the multiple television displays in a typical television sales store front. However, when one needs to make decisions about acceptability of color for a display, one needs to know with a great degree of confidence that the color, that one is observing and discussing with an operator who is observing the same image on a different CRT, is the same as the color observed by the operator. The same is true, if the displayed image is one created on a printer and compared with the same image created on a different printer.

Colorimetry, which is the study of color based on both spectral distribution of the energy reflected or transmitted from a test sample and the response of the human eye, as well as the spectral distribution of the illuminating source, provides a method to describe and measure color and enables one to determine when colors match. Through the use of CIE defined Tristimulus Values (which are the amounts of three primary lights which when added produce a visual, or colorimetric match with an original color), one may determine with reasonable certainty that if two colors have the same three CIE Tristimulus values that is, if the Red Tristimulus value of one color is the same as the Red Tristimulus value of the other, and so on for the Green and Blue Tristimulus values, then the appearance of the two colors will be the same to the average observer. CIE stands for the International Commission on Illumination.

It appears, therefore, that the problem of matching color outputs of different displays is readily resolved by providing displays having the same CIE tristimulus values for the same image value inputs.

While the solution in principle is simple, creating a conversion device which will make two or more different displays produce the same tristimulus output values for the same input image values is very difficult. Each of the display devices operates in its own, device dependent, color space where image values at its input are transformed into display image values. The difficulty lies primarily in matching the two transformations occurring within the two display devices for the image values appearing at the input of each, so that both display the same displayed colorimetric values for the same input image values.

The prior art solutions to color matching fall into two fundamentally distinct approaches. The first is based on the decomposition of a color vector to a set of primaries, and known as the primary decomposition technique. Typical of this approach is the technique described in U.S. patent application 5,196,927 filed Mar. 16, 1990, assigned to E. I. du Pont de Nemours and Company. According to this technique, the input image is decomposed into the linear combination of eight primaries (yellow, magenta, cyan, black, red, green, blue, and three-color black). Coefficients for a 4.times.8matrix (the values of CYMK for each of the eight primaries) are adjusted such that the color of the eight primaries in the input from both systems to be matched are matched, and the new input values are found by matrix multiplication of an 8.times.1 matrix (coefficient of decomposition) with the 4.times.8 matrix. This technique suffers because of the non linearity in the additive properties of color dyes and because the "primaries" used are not mathematically independent.

The more common solution, adopted by the printing industry is the grid sampling technique. This involves using a color transformation formula such that the error between the targeted color and the processed color is minimized. The differences between the various methods in existence are found in the specific transformation formulas. U.S. Pat. No. 4,500,919 issued to Schreiber is a good example of the use of a transformation formula. The difficulty with this approach again lies in the non-linearity of the color addition process and the complexity of the color surfaces in a set of equations.

SUMMARY OF THE INVENTION

The present invention relates to provide a method for matching the color display of at least a first and a second display devices, comprising:

(I) creating a transform LUT for converting input color values to output color values by:

(1) producing a first preselected plurality of input color values;

(2) using said plurality of input values to display a plurality of color patches in said first and said second display devices;

(3) obtaining a colorimetric value of each of the displayed color patches in each of the displayed devices and using said colorimetric values to create a first and a second model for the first and second devices respectively correlating preselected input color values to displayed colorimetric values for each of the two devices;

(4) inputting to said first and second models a second preselected plurality of color values;

(5) comparing the output of the first model to the output of the second model to obtain an error signal indicative of the difference between the two output signals;

(6) using the error signal to modify the input color values to the second model and again comparing the output of the first model to the output of the second model to obtain a new error signal;

(7) repeating the process of steps (5) and (6) above until the error signal is a minimum; and

(8) using the modified color values to create a transform LUT correlating input values to modified values;

(II) using the transform LUT to transform any plurality of source color image values before inputting said image values to an input of the second display device; and

(III) displaying said transformed source color image values on said second display device.

When there is no corresponding source image value in the transform LUT correlating an input source color image value to a modified color image value, one can use interpolation to derive a corrected source image value from a closest source and corresponding modified value in the LUT.

Preferably, the first and second preselected pluralities of color image values are the same.

This invention further relates to an apparatus for generating a transform LUT for converting input color values to output color values comprising:

(1) digital color image values input means;

(2) first and second display model means for producing each an output colorimetric tristimulus image values for input color values, each of said model means having an input and an output, the second model input connected to the input means;

(3) an adder device having a first signal input connected to the input means, a second, correction error signal input, a control signal input, and an output connected to the first model input, for outputting modified color values;

(4) means connected to the output of said first and second model means, for comparing the tristimulus values output of the first and second display models and for producing an error signal;

(5) means for testing the error signal to determine if said signal is a minimum error signal, and for outputting a correction error signal and a control signal to the adder, said testing means connected between said means to compare and said adder device; and

(6) means also connected to the input means, the adder output, and the means for testing, for receiving the adder output and for generating a transform LUT correlating digital input color values to the modified color values appearing at the output of the adder when the means for testing determines that the error signal is a minimum.

The apparatus may further comprise means for storing the generated LUT. A gamut mapping means may be included between the output of the second model and the means to compare for mapping output values from the second model into a color gamut commensurate to that of the first model.

The apparatus may comprise hardware, or may be a computer programmed through software to perform all of the above operations, or may be a combination of dedicated hardware and computer implemented software.

BRIEF DESCRIPTION OF THE DRAWINGS

The invention can be more fully understood from the following description thereof in connection with the accompanying FIGURES described as follows.

FIG. 1 is a schematic representation of an arrangement in which two matched images are displayed on two CRT displays.

FIG. 2 is a schematic representation showing the arrangement used in obtaining data for generating the LUTs used in modeling the two displays of FIG. 1.

FIG. 3 is a schematic representation of the process for obtaining the transform LUT used in the adaptor in the arrangement shown in FIG. 1.

FIG. 4 is a generic representation of a display model.

FIG. 5 is a schematic representation of the process used in deriving the correction factors shown in FIG. 3.

FIG. 6 is a schematic representation of an arrangement in accordance with this invention in which two matched images are printed using two printing devices.

FIG. 7 is a schematic representation showing the arrangement used in obtaining data for generating the LUTs used in modeling the two printing devices of FIG. 6.

FIG. 8 is a schematic representation of the process for obtaining the transform LUT used in the adaptor in the arrangement shown in FIG. 6.

FIG. 9 is a schematic representation of an arrangement in accordance with this invention in which two matched images are printed using a CRT display and a printer including an RGB to CMYK converter.

FIG. 10 is a schematic representation of an arrangement in accordance with this invention in which two matched images are printed using a CRT display and a color proofing device using a YMCK to RGB to drive the CRT display.

FIG. 11 is a generic representation of a display model for use in developing the transfer LUTs required for the embodiments shown in FIGS. 9 and 10.

DESCRIPTION OF THE PREFERRED EMBODIMENT(S)

The invention will next be described in detail with reference to the drawings in which similar characters indicate similar elements in all figures of the drawings. During the following discussion, color values are quantized in 256 steps corresponding to an 8 bit system. Other quantizations are possible, and not intended to be excluded because of the use of the 256 steps in the following examples.

The invention comprises a method and apparatus for providing color visual matching of two representations of the same image when the image is displayed in two different displays, so that an observer will on visual examination of the displayed images observe two images having substantially the same color appearance. Using the method or apparatus of this invention causes each of two observers, whose visual response substantially conforms to the visual response of an average observer as defined by the CIE institute, upon observing any one of the color displays under the same or similar surrounding illumination and background, to receive substantially the same visual impression.

FIG. 1 shows in schematic representation a situation where a color image is displayed on two CRT type displays which receive RGB type inputs to display an image. The image may be stored and manipulated in digital format in a work station 10 which itself may comprise a scanner, an image processor, a display device, and have image storage capabilities, or any portion of the above, or more.

Because the required input to displays 12 and 14 in this example is a digital RGB type signal, the work station outputs over line 16 a digital RGB signal representing a color image. The digital RGB signal comprises a set of color values, one set of three values (R, G, B) for each picture element of the image to be displayed. The signal is directed over line 18 to the first display monitor 12 on which the colored image is displayed. The digital RGB signal is also sent over line 20 to the second display monitor 14, which could be located at a location different than the location of the first display 12.

Prior to applying the RGB signal to the second monitor 14, there is interposed an adaptor 22 which receives the digital RGB signal and after processing the signal, outputs a signal comprising a new set of digital color values R', G', B' over line 24 to an input of the second display monitor 14 for each image pixel.

The adaptor 22 comprises a receiving and determining element 26 which receives each set of R, G, B values and checks them against a plurality of sets of R, G, B values comprising a transform look up table (LUT) 28 which correlates R, G, B values to a new set of R', G', B' values which when used as an input to the second display monitor 14 will generate a pixel whose colorimetric value will match that of the pixel generated by the same R, G, B set of color values in the first display monitor 12.

If the receiving and determining element 26 identifies a corresponding set of R, G, B values in the LUT 28, it outputs the appropriate R', G', B' values on line 24. If not, the R, G, B input is directed to an interpolator 30 where R', G', B' values are derived by interpolation using existing adjacent sets of R, G, B values in the LUT. Preferably, the selection of the R, G, B values for the development of the transform LUT 28 is such that values derived by interpolation produce a visually acceptable match of the displayed pixel in the two monitors 12 and 14.

For this system to operate with any degree of success, the development of the transform LUT 28 is critical. The values generated by this LUT 28 must indeed lead to values that correctly reproduce colors on the second display 14 which visually match the colors on the first display 12. FIGS. 2, 3 and 4 help explain how the LUT 28 is generated.

The first step in developing the transform LUT 28 is the creation of two models representing the two displays whose output is to be matched. FIG. 2 shows how this is accomplished.

The work station 10 is used to generate and output over line 16 a set of digital R, G, B color values in a regular pattern which is selected to supply color values closely enough spaced so that interpolation of values between existing values is reasonably accurate. Preferably, but not necessarily, the output is adjusted so that the color values produce displays 32 and 34 on the first and second display monitors 12 and 14, respectively, comprising a plurality of different color patches on display screens of the display monitors 12 and 14. As an example, the work station 10 may generate a plurality of digital RGB values such as shown in the following table I:

TABLE I ______________________________________ R: 0, 13, 26, 51, 76, 102, 128, 153, 178, 204, 230, 255. G: 0, 13, 26, 51, 76, 102, 128, 153, 178, 204, 230, 255. B: 0, 13, 26, 51, 76, 102, 128, 153, 178, 204, 230, 255. ______________________________________

Each combination of R, G, B values from this table represents a set (R, G, B).sub.n of color values. In this illustration where there are 1728 possible such sets of R, G, B values, n=1 to 1728.

These 1728 sets of R, G, B values are supplied to both display monitors 12 and 14 over lines 18 and 20' creating a total of 1728 displayed patches for each of the monitors 12 and 14, corresponding to the 1728 different combinations of the selected values. All 1728 patches are not necessarily displayed on each monitor screen simultaneously.

A colorimetric measuring device 36 is used to read the displayed patches displayed in the first and second monitors 12 and 14 as shown in FIG. 2. The output of the colorimetric device is a colorimetric set of color values for each of the patches. In our preferred embodiment, the output of the colorimeter are CIE defined L, a and b sets of color values in the Lab color space. Output colorimetric measurements given in other colorimetric color spaces are equally acceptable, such as XYZ, Tristimulus CIE defined RGB, etc. Uniform as well as non-uniform color spaces may be used. However, values in non-uniform color spaces related to a uniform color space with a known mathematical relation may result in increased calculation steps which must be compared with any gained advantage as a result of this selection, to determine if the particular choice of color space is justified. (For a description of the different color spaces and related terminology, see, in general, DIGITAL IMAGE PROCESSING, 2nd edition, by William Pratt, published by John Wiley and Sons, Inc. pages 62-73.)

Once the measurement of all patches is completed, two LUTs are compiled, representing the transfer functions of the two monitors 12 and 14. The first LUT will consist of the (RGB).sub.n sets of values and the corresponding (Lab).sub.n sets of values read off the display on monitor 12 and the second LUT will consist of the same (RGB).sub.n sets of values and the corresponding (L'a'b').sub.n sets of values read off the display monitor 14.

Referring to FIG. 3, a first and a second model 40 and 42 are built representing the two monitors 12 and 14, respectively. FIG. 4 shows a generic model 44 structure, which is used throughout this description in this invention to convert color values. The model 44 comprises an LUT 48, which is the particular device derived LUT representing the device transfer function. In this example, the LUT 48 for the first model 40 will be the first LUT correlating the (RGB)n sets of color values to the (Lab).sub.n sets of values, and the LUT 48 for the second model 42 will be the second LUT correlating the same (RGB).sub.n sets of color values to the (L'a'b')n sets of values.

In addition to the LUT 48, the model 44 comprises a receiving and determining means 46 similar to, or the same as, the receiving and determining means 26 described earlier, and a mathematical interpolating means 50, also similar to, or the same as, the previously described interpolating means 30.

FIG. 3 schematically represents the generation of the transform LUT. Sets of R, G, B color values from a preselected plurality of R, G, B sets of color values are used. This plurality of R, G, B sets of values is, preferably, also produced in the work station 10, and, again preferably, is the same as the sets of (R, G, B).sub.n values from Table I previously used to generate the two LUTs for the two display models 40 and 42.

Each (R, G, B) n set of values is directed to the input of the first model 40 over lines 54 and 58. Model 40 produces an output of (Lab)n values corresponding to the input (R, G, B).sub.n values for this model 40. The same (R, G, B).sub.n values are directed to the second monitor model 42 over line 56. Ahead of model 42, there is an adder 60 which operates to add to the R.sub.n, G.sub.n, and B.sub.n components of the (R, G, B).sub.n set of values any correction factor dR.sub.n, dG.sub.n, dB.sub.n, appearing on line 78. At first, nothing appears on line 78, and the R.sub.n, G.sub.n and B.sub.n values are applied to the input of model 42 unaltered. Model 42 also produces an output set of (L'a'b').sub.n color values corresponding to the input (R, G, B).sub.n. This output appears on line 64.

The (Lab).sub.n values over line 72 and the (L'a'b').sub.n values over line 64 are next compared in comparator 66 and a difference signal [(Lab).sub.n -(L'a'b').sub.n ] is generated and directed over line 74 to correction factor generator 76. Using this difference signal, correction factor generator 76 produces correction factors dR.sub.n(1), dG.sub.n(1), dB.sub.n(1) (in a manner to be explained herein below) which are sent over line 78 and added to the corresponding components of the (R, G, B).sub.n set of values in adder 60, resulting in a new set of color values R.sub.n(1) '=R.sub.n +dR.sub.n(1), G.sub.n(1) '=G.sub.n +dG.sub.n(1), and B.sub.n(1) '=B.sub.n +dB.sub.n(1) on line 62. These new values are again applied to the input of model 42 which produces a new output set of L.sub.2 'a.sub.2 'b.sub.2 ' values on line 64 which is again compared in comparator 66 with the set of (Lab).sub.n values from model 40. A new set of correction factors dR.sub.n(2), dG.sub.n(2), and dB.sub.n(2) is produced and added to R.sub.n(1) ', G.sub.n(1) ' and B.sub.n(1) ' to produce a new set of R.sub.n ', G.sub.n ', and B.sub.n ' values An error "E" defined as:

E=[(L--L').sup.2 +(a--a').sup.2 +(b--b')2].sup.1/2

is used to determine when to stop this cycle. Referring to FIG. 5, error "E" is compared in comparator 92 with preselected minimum acceptable limits or a "0" value. If the error is "0", or within the preselected acceptable limits, a signal over line 94 switches switch 98 (which is part of adder circuit 60) to feed the R.sub.n ', G.sub.n,' and B.sub.n ' values which produces the minimum over line 80 to the transform LUT compiler 82, rather than to model 42; the next input set of values (R, G, B).sub.n+1 is then applied over line 54 to the two models and the whole process is repeated for this new set of values and so on.

The set of R.sub.n 'G.sub.n 'B.sub.n ' values which produced this zero or acceptable minimum error are sent over line 80 to the LUT compiler 82 where the (R, G, B).sub.n and corresponding (R'G'B').sub.n sets of values are used to compile the transform LUT 28 so that for each set of (R, G, B).sub.n values there is a corresponding set of (R',G',B',).sub.n values. Means to store the transform LUT 28 is included in the compiler 82. This LUT 28 is the the same transform LUT shown in FIG. 1 discussed earlier.

The correction factor generator 76 produces the dR, dG and dB correction factors as follows. The input to the correction factor generator 76 is the difference signal produced by the comparator 66, specifically [(Lab).sub.n -(L'a'b').sub.n ]. The individual components of this signal are: dL which equals (L--L'), da which equals (a--a') and db which equals (b--b'). The correction factors dR, dG and dB, and the difference signals dL, da, and db are related by the following relationships: ##EQU1##

The actual numeric values for .differential.L/.differential.R, .differential.L/.differential.G, .differential.L/.differential.B, etc. for use in solving the above system of equations, are obtained by the process schematically illustrated in FIG. 5. The input R.sub.n, G.sub.n,B.sub.n values are changed by holding two the same and incrementing the third by one unit in incrementing module 96. This module comprises a memory for temporarily holding the input values of R.sub.n, G.sub.n, and B.sub.n and an arithmetic means for incrementing each of those values by 1 and for applying various combinations of R.sub.n, G.sub.n, B.sub.n, R.sub.n +1, G.sub.n +1, and B.sub.n +1 sets of values to the input of model 42.

An example which uses illustrative RGB values, will be used to explain the derivation of the numerical values for the partial derivatives needed to solve the above equations, (1) (2) and (3). Let the input R.sub.n, G.sub.n, and B.sub.n values for a set of (R, G, B).sub.n values be 100, 100, 100. Assume that when these values are applied to both models 40 and 42, two sets of Lab values, (Lab).sub.n and (L'a'b').sub.n are produced such that the error E as previously defined is different than "0" or a preselected acceptable minimum. Incrementing module 96 produces over line 65 three new color value sets, R.sub.n, G.sub.n, (Bn+1), having the numerical values (100, 100, 101) in this illustrative example, R.sub.n, (Gn+1),Bn, having the numerical values (100,101,100) and (Rn+1), Gn, Bn, having the numerical values (101,100,100). These new values are applied to model 42 over line 63'. The resulting variations in the components of the (L'a'b') n values produced by model 42 are directed to an arithmetic calculator 82 over line 75 and to buffer memory 84. The calculator 82 performs the operations: (L--L')/R-(R+1), (L--L')/G-(G+1), (L--L')/B-(B+1), (a--a')/R-(R+1), (a--a')/G-(G+1), (a--a')/B-(B+1), (b--b')/R-(R+1), b--b'/G-(G+1) and b--b'/B-(B+1), to derive the partial derivative values (.differential.L/.differential.R, .differential.L/.differential.G, etc.) used in equa