JPH10243247A - Color information conversion method and device to match observation results of colors with each other - Google Patents

Abstract

PROBLEM TO BE SOLVED: To realize highly accurate conversion between a color separation value and a spectral reflectance by employing the spectral reflectance, independent of a light source and effectively utilizing a learning function of a neutral network. SOLUTION: A control section 21 uses a 5th control means to obtain a color- matched color separation value through optimization under the condition of minimizing a square error (square norm) or a mean color difference between a color under a prescribed light source that is calculated by applying a spectral reflectance to an output from a 2nd control means 23 and a color under a prescribed light source, that is outputted from a 4th control means. A memory 24 stores a combination coefficient of a neural network, whose learning is finished, spectral distribution of the light source possibly in use and a color- matching function or the like. The combination coefficient for a 3-layer neural network has been learned and corrected in advance, so that a square error between an output of each output unit and a spectral reflectance given as a teacher signal is minimized.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a color DTP
Due to the influence of the surrounding environment (illumination conditions, background color, etc.) on images displayed on image output devices such as color printers and displays in skTop Publishing), images with the same color value look different. Regarding the color matching method and apparatus in such a case, in particular, a spectral reflectance that accurately represents physical (spectral colorimetric) characteristics of a color is used as an intermediate expression of a color, and furthermore, a color appearance model is used. By using the conversion between the color value and the spectral reflectance determined so as not to depend on the surrounding environment represented, the colors displayed on the image output device can always be seen to be the same even in different viewing environments. Can be modified,
In addition, the present invention also relates to a color information conversion method and apparatus for matching the appearance of colors that can accurately convert color information under any light source.

[0002]

2. Description of the Related Art Normally, when an object is viewed by the human eye, the human eye adapts to an observation environment such as peripheral illumination or a background color near an image (the spectral sensitivity characteristics of a photoreceptor in an eyeball change). )
Therefore, if the viewing environment is different, the color of the image looks different even when the same image is viewed. Therefore, in order to make the colors of an image look the same under different viewing environments, it is necessary to appropriately adjust the colors of the images displayed / output to the image output device according to the viewing environment. On the other hand, recently, color DTP (De
sk-Top Publishing) will be commonly used for personal publishing and in-house printing, and it will be possible to edit and modify digital images as they are on a computer, create color documents and computer graphics. Therefore, the difference in color appearance depending on the observation environment has become a serious problem.

For example, when comparing an image displayed on a display with an image output by a printer, the background color of the display differs from the color of the paper, and the brightness of the display and the room in which the image printed on the paper is observed. Because the brightness is different, the adaptation of the eyes when comparing each other is different, so even if the colors are colorimetrically matched,
The color appearance of the image will be completely different. Recently, a color appearance model (Color Appearacne Model) has been developed as a technique for correcting differences in color appearance due to differences in viewing environments.
There is a method of predicting the state of adaptation of the human eye in a predetermined observation environment using an image, and adjusting the color displayed on the image output device based on the prediction (MD Fairchild: Considering the
Surround in Device-independent Color Imaging, Col
or research and application Vol.20, No.6 (1995)).

[0004]

On the other hand, in order to properly adjust the colors displayed on the image output device, it is necessary to convert between the color values and the color separation values specific to the image output device. Many methods have been proposed for the conversion, such as a method using a lookup table and a method using a neural network.However, since the color value is a value that depends on the light source, the conventional color value In the conversion using the conversion between the color separation values, one conversion system is required for one light source.
Therefore, in order to correct the difference in color appearance, it is necessary to create an enormous number of conversion systems for all possible light sources and image output devices, which is practically impossible. There is a problem that it is not possible to actually construct a system that can correct the difference in color appearance under (light source).

[0005] As described above, the conventional technique for correcting the difference in color appearance is based on an experiment for correcting the difference in color appearance under a limited number of observation environments set in advance. It is difficult to create a practical system that can correct the difference in color appearance under an arbitrary observation environment (under a light source) as actually required . The present invention has been made in view of the above-mentioned circumstances, and color information that matches the appearance of colors so that color information values under an arbitrary light source can be converted into color separation values regardless of the type of observation environment (illumination). It is an object to provide a conversion method and device.

[0006]

According to the present invention, there is provided a color information conversion method and apparatus for matching a color appearance, which comprises a method for matching a color appearance of a reproduction device observed under different environments, with a spectral reflectance or A method of converting one of the spectral transmittances into a color separation value, wherein one of the spectral reflectance or the spectral transmittance is set as an intermediate color system of a color, and the color separation value is used as the intermediate color system. A first conversion step of converting at least three color separation values into the intermediate color system using a neural network trained to convert to the intermediate color system;
A first calculation step of obtaining a color value at a predetermined light source from the intermediate color system, a second calculation step of converting the color value into a color value that does not depend on the surrounding environment to be observed, A third calculation step of calculating a color value in a predetermined environment from a color value not to be used, and a step of optimizing a color value converted by a neural network from the color value and a predetermined color separation value. Features.

That is, any one of the spectral reflectance or the spectral transmittance independent of the light source is used as an intermediate color system, and conversion from a set of color separation values to either the spectral reflectance or the spectral transmittance is performed. Is converted from the color separation value set to either spectral reflectance or spectral transmittance using a nonlinear transformation by a neural network that has been trained, and the spectral reflectance or spectral obtained by the conversion is converted. The color value at the specified light source calculated from the transmittance is converted into a color value independent of the viewing environment using a color appearance model, and further converted into a color value under a predetermined viewing environment. The relationship between a color value under a predetermined observation environment and a color reflectance value calculated from one of a spectral reflectance or a spectral transmittance obtained by a neural network from at least three color separation values. Optimization under constraints that, so that to realize an image of a color correcting to be displayed on the image output device so see the color match under a given viewing environment.

Here, the color separation values are three primary colors of additive color mixture such as R, G, and B values constituting a color image or three primary colors of subtractive color mixture such as C, M, Y, and K values constituting a color image. And Further, the color values are X, Y, and Z tristimulus values defined by the CIE as color values, or R,
G and B values, and L representing the amount of response of the cone of the human eye,
M and S values are set. Further, the two neural networks are multilayer feedforward type neural networks having three to five layers, and the neural network has a learning function.

Further, as the constraint condition, a color value under a predetermined light source obtained by a color appearance model and a color value at a specified light source calculated from a spectral reflectance obtained by the neural network are used. The condition is that the square error (square norm) between them or the average color difference is minimized by an optimization method. In addition, a non-linear optimization method, which is a general method of extremum search, is used as the optimization method. The predetermined light source is CIE
A light source, B light source, C light source, and D of any color temperature specified by
Light sources, F1 to F12 light sources, and light sources having an arbitrary light source spectral distribution obtained by measurement with a spectral colorimeter are used.

Further, according to the present invention, there is provided a color information conversion apparatus for matching color appearance, comprising: input means for inputting at least three color separation values; image output means for outputting an image from the color separation values; First conversion means for converting the set of separation values into spectral reflectance by a multilayer feedforward type neural network learned according to a predetermined image output characteristic; and converting the set of color separation values input to the input means into the image A second conversion unit for converting into a spectral reflectance or a spectral transmittance by a neural network learned according to the characteristics of the output unit; and a spectral reflectance or a spectral transmittance determined by the first conversion unit. Calculate a color value independent of the surrounding environment calculated by a color appearance model from a color value in a predetermined environment calculated using one of the above. Third conversion means, fourth conversion means for calculating a color value under a predetermined environment from color values irrelevant to the surrounding environment obtained by the color appearance model, and output from the fourth conversion means. So as to minimize at least one of the square error and the average color difference between the color values under the predetermined environment calculated from the color values under the predetermined environment and the spectral reflectance output from the second conversion means. Optimizing means for optimizing the input color separation value.

That is, at least three color separation values are converted into spectral reflectances, and furthermore, the color appearance under an arbitrary observation environment is adjusted by using a color value calculated using a color appearance model. In the color information conversion imaging device, input means for converting the at least three color separation values into an electric signal or an optical signal; input means for converting at least three color separation values into an electric signal; First converting means for converting an electric signal into one of spectral reflectance and spectral transmittance by a neural network learned in advance and outputting the converted signal as an electric signal, and an electric signal corresponding to an initial value of an appropriate color separation value A second conversion means for converting the signal into one of spectral reflectance and spectral transmittance by a neural network trained in advance and outputting it as an electric signal; The coupling coefficient of the neural net,
And storing the CIE color-matching function or the R, G, B color-matching function, the spectral sensitivity characteristic of the photoreceptor (cone) of the human eye, and the spectral distribution of the light source that may be used. After calculating a color value under a predetermined light source from the output of the storage device and the first conversion means, and converting the color value into a color value independent of the surrounding environment using a color appearance model A third converter for outputting as an electric signal, a fourth converter for converting a color value independent of the surrounding environment into a color value under a predetermined environment, and outputting the converted signal as an electric signal;
Optimization under a constraint on the relationship between the conversion means and the color value under a predetermined environment calculated from one of the spectral reflectance and the spectral transmittance output from the second conversion means. Optimizing means for converting color separation values adjusted according to the above, and output means for outputting an electric signal output from the optimizing means as an arbitrary output signal, wherein the first and second converting means are 3 to 5 A neural network having a plurality of layers, said neural network having a learning function, and converting at least three color separation value sets into a spectral reflectance or a spectral transmittance of a reproduced color corresponding to the color separation value set. Each of them is trained to convert to a rate distribution, and also a color appearance model in a fourth conversion unit and a color matching function stored in a storage device, or a spectral sensitivity of a cone. Using the color value under a predetermined environment calculated using the characteristic, the spectral reflectance output from the second conversion unit and the color matching function stored in the storage device, or the spectral sensitivity characteristic of the cone Under the condition that the square error (square norm) between the calculated color value and the color value under the predetermined environment or the average color difference is minimized, the input of the second conversion unit is optimized using an optimization method. The optimization realizes color information conversion for correcting a difference in color appearance.

Therefore, in the color information conversion apparatus for matching colors according to the present invention, by effectively utilizing the learning function of the neural network, even when a color is expressed by a CMY value or a CMYK value, the RGB value can be obtained. Even when the expression is expressed by, the conversion between the color separation values and the spectral reflectance for these colors can be realized with high accuracy. Further, since a neural network having a learning function is used for conversion between a color separation value such as a CMY value, a CMYK value, or an RGB value and a spectral reflectance or a spectral transmittance, a sufficiently learned neural network is used. Due to its generalization ability, the network can obtain an appropriate output for the input data even when unknown data not used for learning is input.

Further, using a model of color appearance, a color value under a predetermined light source, which is an output, and a color value under a specified light source calculated from a spectral reflectance obtained by a neural network. By using a non-linear optimization method that restricts minimization of the square error (square norm) or the average color difference, a color that matches the appearance of highly reliable colors independent of the device and the light source. Information conversion can be realized.

[0014]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS First, a basic concept of a color information conversion apparatus according to the present invention for matching color appearance will be described. Color is colorimetrically defined as the value obtained by integrating the product of the spectral distribution of the light source, the spectral reflectance of the object, and the color matching function in the entire visible light region, that is, the color value. According to this definition, in order to match the color appearance under different viewing environments, once
Find the color that the object that is unrelated to the observation environment originally has,
Then, it is necessary to predict the color appearance when viewed in different viewing environments, and to output to the image output device a color that causes human vision to cause the color appearance under the environment.

The present invention is based on such a basic idea for matching the appearance of colors, and is applied to color reproduction independent of devices and viewing environments in the field of color reproduction, and different images are obtained under different viewing environments. Achieving matching of color appearance between output devices. The color information conversion apparatus for adjusting the color appearance according to the present invention uses an image output unit and a neural network that has learned conversion between an input color separation value and spectral reflectance. Further, the conversion between the viewing environment-dependent color value and the viewing environment-independent color value, and the conversion from the viewing environment-independent color value to the dependent color value are performed using a color appearance model. In addition, the image is converted into a color separation value that matches the appearance of a color under a predetermined environment by optimization satisfying a predetermined condition. The optimization includes the color values obtained from the color appearance model,
This includes a square error (square norm) between the color values calculated from the spectral reflectance or a condition of minimizing the average color difference, which will be described in detail later.

Hereinafter, a color information conversion apparatus for matching the appearance of colors according to an embodiment of the present invention will be described with reference to the drawings. FIG. 1 shows a functional configuration of a color information conversion apparatus that matches the appearance of colors according to an embodiment of the present invention. As shown in FIG. 1, a color information conversion device that matches the appearance of a color includes an input unit 1.
0, a color information conversion unit 20, and an image output unit 30. The input unit 10 inputs a color separation value such as a CMY value and outputs the color separation value to the color information conversion unit 20 as a corresponding electric signal. The image output unit 30 outputs a color DTP (DeskTopPublis) that outputs an image on a display or paper according to the input color separation value.
hing), an image output device such as an electrophotographic printer or display is applied. The color information conversion unit 20 includes a control unit 21 that controls the entire color information conversion process for matching the appearance of colors, a first conversion unit 22 and a second conversion unit 23 that are configured by a neural network, and a color value of a predetermined observation environment. A third conversion unit 24 and a fourth conversion unit 25 that perform conversion between color values and color values that do not depend on the viewing environment; a fifth conversion unit 26 that performs conversion processing under predetermined conditions; The memory 27 stores various parameters related to the conversion unit 26.

The first conversion unit 22 and the second conversion unit 23 each have a conversion function described later, and at the same time, a learning process is performed under the control unit 21 so as to realize this function. Further, the third conversion unit 24 converts the color appearance under the designated light source calculated from the spectral reflectance output from the first conversion unit 22 to a color value independent of the viewing environment, as described later. Convert using the model. Further, the fourth conversion unit 25
Converts a color value independent of the observation environment into a color value in a predetermined observation environment (illumination) using a color appearance model. In addition, the fifth conversion unit 26, under the control of the control unit 21,
The square error (square norm) between the color value under the predetermined light source calculated from the spectral reflectance output from the second conversion unit 23 and the color value under the predetermined light source output from the fourth conversion unit 25, or Optimization is performed under the condition of minimizing the average color difference, and color-matched color separation values are obtained.

The second converter 23 converts an electric signal of an initial value having an appropriate value into a spectral reflectance. FIG. 2 shows an example of a neural network used in the first conversion unit 22 and the second conversion unit 23. As shown in the figure, the intermediate layer may be composed of neural circuit elements whose input / output relationship is represented by a sigmoid function, and the input / output layers may be linear neural circuit elements. The output layer is desirably formed of a neural circuit element having a sigmoid characteristic similarly to the intermediate layer. The neural network shown in FIG. 2 has output units U1, U2,
... Un, the number of units in the input layer is 3, the number of units in the intermediate layer is 18, and the number of units in the output layer is 31. However, it is desirable that the number of intermediate units be determined according to the characteristics of the image output device used.

The control section 21 uses a fifth conversion section 26 to calculate a color value under a predetermined light source calculated from the spectral reflectance output from the second conversion section 23 and a predetermined color value output from the fourth conversion section 25. A color-matched color separation value that is optimized under the condition of minimizing the square error (square norm) of the color value under the light source or the average color difference is obtained. The memory 27 stores the coupling coefficient of the neural network for which learning has been completed, the spectral distribution of the light source that may be used, the color matching function, and the like. Next, the first conversion unit 22 and the second conversion unit 23 configured by a neural network will be described in detail.

The input unit 10 shown in FIG. 1 converts at least three sets of color separation values for colors into corresponding electric signals. That is, R (red), G (green), B (blue), C (cyan), M (magenta), and Y (yellow) of the original colors to be subjected to color information conversion.
It is output as an electric signal indicating a value or a color separation value such as C (cyan), M (magenta), Y (yellow), and K (black) values. In the first conversion unit 22 and the second conversion unit 23, when color separation values such as RGB values, CMY values, and CMYK values are given to the input layer as appropriate initial values, a neural network having a structure as shown in FIG. Is output according to the following equation.

[0021]

(Equation 1)

Here, W (h) ij represents the coupling coefficient between the j-th input unit and the i-th intermediate unit, Ii represents the output of the intermediate unit, and bj represents the bias value. The output Oi of the output layer unit is similarly output according to the following equation.

[0023]

(Equation 2)

Where W (o) ij is the j-th intermediate unit and i
Represents the coupling coefficient of the ith output unit, and Hi is the output value of the ith intermediate unit defined by the above equation (1). Also, n, m, and k are an input layer, an intermediate layer,
This is the number of units in the output layer. At this time, the characteristics of the unit in the input layer are such that the input is output as it is, and the function f (x) of the hidden unit is defined by the following equation.
This is a characteristic represented by a sigmoid function that monotonically increases between [0, 1].

[0025]

(Equation 3)

The output unit has a characteristic represented by a sigmoid function or a linear function. However, the output layer may have an input from the bias unit. The coupling coefficient of the three-layer neural network is previously learned and corrected so as to minimize the square error between the output of each output unit and the spectral reflectance given as a teacher signal. The learning uses the back propagation learning rule proposed by Rumelhert, and learns the conversion from color separation values to spectral reflectance. Then, the coupling coefficient of the learned three-layer neural network is stored in the memory 27. On the other hand, the third conversion unit 23
Then, the spectral reflectance R, which is the output of the first conversion means 22 obtained by the neural network after learning,
(C, M, Y), color matching function M, spectral distribution L of specified light source

[0027]

(Equation 4)

The color values calculated in step (4) are calculated, and the color values dependent on illumination obtained by the above equation (4) are converted into color values independent of the viewing environment using a color appearance model. Here, the RLAB (Mark, D. Fairchild, Roy S. Berns:
Image Color-Appearance Specification ThroughExten
sion of CIELAB, Color research and Application Vol
ume 18, Number.3, June 1993). In this model, the type of observation illumination, background brightness, color, and the like are used as the observation environment. Fourth converter 25
Then, from the color value independent of the observation environment, which is the output value of the third conversion unit 24, to a color value under an observation environment different from the observation environment used in the conversion of the third conversion unit 24, RLA
Convert using B.

In the fifth converter, a color value under a predetermined observation environment, which is an output value of the fourth converter, and a predetermined light source calculated from a spectral reflectance, which is an output value of the second converter, are used. The input of the second conversion unit is optimized under the condition of minimizing the square error (square norm) between the color value and the average color difference, and converts the color value under a predetermined environment into a color. Is converted to a color separation value that has been subjected to processing for visual matching. The converted color separation value set is sent to the image output unit 30 and is converted into an arbitrary signal such as an electric signal or an optical signal. Next, this first
The learning process of the first conversion unit 22 and the second conversion unit 23 in the embodiment will be described.

FIG. 4 shows a learning process of the first conversion unit 22 and the second conversion unit 23 constituted by a neural network. As described above, this learning process is performed under the control of the control unit 21. The spectral reflectance data used for learning of the neural network includes a target image output device and a color chip of 1331 colors printed by changing CMY values from 0 to 100% at intervals of 10% using a spectral colorimeter. Visible light region from 400nm to 700nm obtained by coloring is 10
Data of 31 points sampled at nm intervals are used, and CMY values of a color chip are used as color separation values. The learning data uses only 216 colors of the data.

First, a color separation value is input to the input layer of the first conversion unit 22 (step S10). Here, the control unit 21
Is input. Thereafter, the spectral reflectance measured by the spectral colorimeter is given to the output layer as a teacher signal (step S11). The CM
The Y values are generated so as to correspond to the teacher data, respectively. Next, a learning process is performed based on the back propagation learning rule (step S12). That is, the coupling strength (coupling coefficient) of the units in the neural network is changed so that the value output from the output layer approaches the spectral reflectance as the teacher signal.

Thereafter, it is determined whether or not the processing has been performed for all of the 216 color patches.
10 is executed again (step S1).
3, NO). If all 216 sheets have been processed, the process proceeds to the next step (step S13, YES). After this,
It is determined whether the sum of the values output from the output layer and the 216 color chips of the spectral reflectance as the teacher signal is within a predetermined error (step 13). If the predetermined learning condition is satisfied, the coupling coefficient of the neural network is stored in the memory 27, and the learning is completed (step 13,
YES). If the learning condition is not satisfied, the learning process from step S10 is executed again from the first color chip (step S13, NO). The second conversion unit 2
The learning process of No. 3 is performed by the same process.

Next, the color misregistration correction processing in this embodiment will be described with reference to the flowchart of FIG. First, the color separation value of each pixel of the image is sent from the input unit 10 to the color information conversion unit 20 (Step S20). At the same time, a random number is input as an initial value in the second converter 23. Here, the values of C, M, and Y of the CMY values are from 0 to 1.
The range is up to 00. Thereafter, the spectral distribution of the specified light source L1, color matching function,
The coupling coefficient of the neural network determined by the learning process is set in the first conversion unit 22 (Step S
21). At the same time, the spectral distribution of the specified light source L2, the color matching function, and the coupling coefficient of the neural network determined by the learning process are set in the second conversion unit 23 (Step S22). In the first conversion unit, the CMY values input from the input unit 10 are converted into spectral reflectances, and the color values of a predetermined light source are sent to the third conversion unit as an output (step S2).
3). Further, the observation environment (spectral distribution of peripheral illumination, chromaticity value of background color, etc.) stored in the memory 27 in advance is read into the third and fourth conversion units (step S24). In the third conversion unit, the color values sent from the first conversion unit are converted into color values independent of the viewing environment by using a color appearance model (RLAB) (step S25). The fourth conversion unit converts a color value independent of the viewing environment, which is an output value of the third conversion unit, into a color value dependent on a different viewing environment from the input of the third conversion unit (step S26). Then the second
The conversion unit converts the CMY values given as the initial values into the spectral reflectances (Step S27).

Further, in the fifth converter, the spectral reflectance, which is the output value of the second converter, the color matching function previously stored in the memory 27, the color value calculated from the spectral distribution of the light source L2, and the like. The optimization process is performed using the output value of the fourth conversion unit (step S28). This optimization processing is performed so as to satisfy a condition of minimizing a square error (square norm) between two response amounts or an average color difference.
The CMY values input to the second conversion unit 22 are updated (Step S29). Next, it is determined whether or not the updated CMY values satisfy an optimization condition, for example, a condition that a square error (square norm) or an average color difference is equal to or less than a predetermined value (step S30). If the optimization condition is not satisfied (step S30, NO), the processing from step S27 is performed again according to the updated CMY values. When the optimization condition is satisfied (step S3
0, YES), and transfer the CMY values obtained by this processing to the image output unit (step S31). FIG. 5 schematically shows the flow of the optimization processing in steps S27 to S30.
Here, the BFGS method, which is a general method of nonlinear optimization, is used to optimize the two response amounts.

Through such processing, color information conversion for matching the color appearance is performed.

[0036]

As described above in detail, according to the present invention, by using the spectral reflectance independent of the light source and effectively using the learning function of the neural network, the color is changed to the CMY value or CMYK. When represented by a value,
Even in the case of being represented by RGB values, conversion between the color separation values and the spectral reflectance for these colors can be realized with high accuracy. In addition, since a neural network having a learning function is used for conversion between a color separation value such as a CMY value, a CMYK value, or an RGB value and a spectral reflectance or a spectral transmittance, a sufficiently trained neural network is used. Due to its generalization ability, it is possible to obtain an appropriate output for input data even when unknown data not used for learning is input. Furthermore, by using a model that matches the appearance of colors as the color values, the color values calculated using the spectral reflectance obtained by the conversion of the neural network are once converted into the observation environment (the brightness of the surrounding illumination, Color temperature, chromaticity value of the background color, etc.), and convert it again to a color value in another viewing environment. The color matching that always becomes the same can be realized. Spectral reflectance and CMY independent of the observation environment and device, and independent of the light source
By describing the colorimetric characteristics of the device by converting the values, the light source information (spectral distribution) can be given when calculating the color values, so there is no need to create a conversion system for each type of light source, and any It is possible to actually construct a system that can correct the difference in color appearance under the environment (under the light source).

[0037]

[Brief description of the drawings]

[0038]

FIG. 1 is a block diagram illustrating a schematic configuration of a color information conversion apparatus that matches the appearance of colors according to an embodiment of the present invention.

[0039]

FIG. 2 is a schematic diagram illustrating a configuration example of a feedforward neural network used in a first conversion unit and a second conversion unit according to an embodiment of the present invention.

[0040]

FIG. 3 is a diagram illustrating a process of causing a feedforward neural network in the embodiment to learn conversion from CMY values to spectral reflectance.

[0041]

FIG. 4 is a flowchart showing color information conversion processing for matching colors in the embodiment according to the present invention.

[0042]

FIG. 5 is a schematic diagram showing an optimization process in a color information conversion process for matching the appearance of colors according to the embodiment of the present invention.

[0043]

[Brief description of reference numerals]

 DESCRIPTION OF SYMBOLS 10 ... Input part 20 ... Color information conversion part 30 ... Image output part 21 ... Control part 22 ... First conversion part 23 ... Second conversion part 24 ... Third conversion Unit 25: Fourth conversion unit 26: Fifth conversion unit 27: Memory

Claims (11)

[Claims] In a method for adjusting the appearance of colors of a reproduction device observed under different environments, one of a spectral reflectance and a spectral transmittance independent of a light source is set as an intermediate color system, and the color separation value is calculated from the color separation value. A first conversion step of converting at least three color separation values to the intermediate color system using a neural network that has been trained to convert to the intermediate color system;
A first calculating step of obtaining a color value at a predetermined light source from the intermediate color system, a second converting step of converting the color value into a color value that does not depend on a surrounding environment for observation, A second calculation step of calculating a color value at a predetermined light source from a color value not to be used, and a step of optimizing a color separation value converted by a neural network from the color value and the predetermined color value. A color information conversion method that matches the appearance of characteristic colors. 2. The optimizing step includes: minimizing at least one of a square error and an average color difference between the color value obtained in the second calculation step and the predetermined color value. 2. The color information conversion method according to claim 1, wherein the color separation value is optimized. 3. The color information conversion method according to claim 1, wherein the color value independent of the surrounding environment to be observed uses a color value represented by a color appearance model. 4. The method according to claim 1, wherein the color value is a CIE (Commission Inter).
nationale de I'Eclairage) defines X, Y, and Z tristimulus values, or R, G, and B values, as color values.
2. The color information conversion method according to claim 1, wherein the color information is one of L, M, and S values representing the response amount of the cone of the human eye. 5. The color separation value is one of C (cyan), M (magenta), Y (yellow), and C, M, Y, K (black) values forming a color image. 2. The color information conversion method according to claim 1, wherein the color information is a subtractive primary color. 6. The color separation value is defined as R
2. The color information conversion method according to claim 1, wherein the three primary colors are additive primary colors of (red), G (green), and B (blue) values. 7. The neural network according to claim 1, wherein said neural network in said first and second transforming steps is a multilayer feedforward neural network.
A color information conversion method that matches the appearance of the described color. 8. The color information conversion method according to claim 1, wherein said optimizing step performs nonlinear optimization. 9. The light source according to claim 1, wherein the predetermined light source is one of a C light source specified by CIE and a D light source having a color temperature of 4000 [K] to 7500 [K]. A color information conversion method that matches the appearance of colors. 10. The predetermined light source is CIE (Commission I).
A light source, B defined by nternationale deI'Eclairage)
2. The light source according to claim 1, wherein the light source is any one of a light source, a C light source, a D light source having an arbitrary color temperature, F1 to F12 light sources, and an arbitrary light source obtained by measurement using a spectrophotometer. A color information conversion method that matches the appearance of colors. 11. An input means for inputting at least three color separation values, an image output means for outputting an image from the color separation values, and at least three sets of color separation values which are learned according to predetermined image output characteristics. A first conversion unit that converts the spectral reflectance into a spectral reflectance by a feedforward type neural network, and a multi-layer feedforward type neural network that learns a set of color separation values input to the input unit according to characteristics of the image output unit. A second conversion unit for converting into a spectral reflectance, and a color appearance model from a color value under a predetermined environment calculated using one of the spectral reflectance and the spectral transmittance obtained by the first conversion unit. Third to calculate the color value independent of the surrounding environment to be calculated
Conversion means, and fourth conversion means for converting the color value output from the third conversion means into a color value under a predetermined environment,
Minimizing at least one of a square error and an average color difference between color values under a predetermined environment calculated from the color values output from the fourth conversion means and the spectral reflectances output from the second conversion means; Optimizing means for optimizing the input color separation values as described above.