JP2000078419A - Method and device for processing color image - Google Patents

Abstract

PROBLEM TO BE SOLVED: To convert a first chrominance signal into a second chrominance signal at high speed with exact color reproducibility based on a color conversion model for finding the first chrominance signal from the second chrominance signal in the case of converting the first chrominance signal into the second chrominance signal and to convert the first chrominance signal into the second chrominance signal at high speed with exact color reproducibility even when the color conversion model for finding the first chrominance signal from the second chrominance signal is not continuous in differentiation. SOLUTION: This method is provided with steps (steps S1 and S2) for finding the color conversion model for finding an L*a*b signals from CMYK signals and a step (step S4) for finding the YMCK signals from the L*a*b signals by a non-linear optimizing method not to use the derived function of the color conversion model such as a simplex method, for example, based on the color conversion model.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

The present invention relates to a method for converting a first color signal into a second color signal.
The present invention relates to a color image processing method and apparatus for converting a color signal into a color signal, for example, a color image processing method and apparatus for converting an RGB signal into a YMCK signal.

[0002]

2. Description of the Related Art Conventionally, an image forming apparatus such as a color printer, for example, applies so-called RGB color signals of red, green and blue and L * a * b * color signals which are signals of a device independent uniform color space. 2. Description of the Related Art A color conversion device for converting into a so-called YMCK color signal of yellow, magenta, cyan, and black, which is an image recording signal (device signal) used for forming an image in an image forming apparatus, is mounted.

As such a color conversion device, generally, a matrix operation or a direct lookup table is widely used. The direct look-up table is a table in which a relationship between an input signal value and an output value obtained by multiplying the input signal by a predetermined color conversion coefficient is stored in a ROM (read-only memory) or a RAM (random access memory) in advance. Yes, it is possible to directly convert an RGB color signal or an L * a * b * color signal into a YMCK color signal. When color conversion is performed using this direct look-up table, there is an advantage that the color conversion can be performed at an extremely high speed because substantially no operation time is required. The color conversion coefficient used when creating the direct lookup table is YM in the image forming apparatus.
A color conversion model is created in advance by modeling the relationship between the CK color signal and the YMCK color signal and a colorimetric value obtained by actually measuring the color when an image is output on paper or the like, and the color conversion model is obtained by utilizing this color conversion model. Have been.

For example, in the invention described in Japanese Patent Laid-Open No. 10-70668, a color conversion model constituted by a conversion table for converting CMY color signals obtained by actual measurement into RGB color signals is successively approximated by a Newton method or the like. , A conversion table for converting RGB color signals into CMY color signals is obtained. Also, Japanese Patent Application Laid-Open No. 9-18689
According to the invention described in Japanese Patent Application Laid-Open No. 4 (1999) -2005, the image recording signal CMYK is obtained from the L * a * b * color signal and the black amount K, and the image recording signal CMYK is input, and The neural network type color conversion model to be output is solved using a successive approximation method such as Newton's method, and L * a
The grid points of the direct lookup table for converting the * b * color signal into the image recording signal CMYK are obtained.

[0005] The method used in the above-mentioned JP-A-10-70668 and JP-A-9-186894 is called computer color matching (CCM), and includes a CM signal containing image recording signals CMY or black K.
YK, tristimulus values XYZ in an image at the time of printout by the image recording signal, and L * a * b * in a uniform color space
The image recording signal CMY or CMY
By creating a color conversion model from which colorimetric values XYZ or L * a * b * can be obtained from K, and solving this color conversion model using a successive approximation method such as Newton's method, the colorimetric value X
In this method, the image recording signal CMY or CMYK is obtained from YZ or L * a * b *.

[0006] Regarding CCM, for example, "Basic and Applied Technology of Color Reproduction", supervised by Kazuo Sayanagi, Trikeps, pp. 147-143.
96-97, which is a method proposed more than 35 years ago, is widely used as a method of obtaining a combination of image recording signals for realizing a required color by using a color conversion model. In general, in the CCM, in order to solve a color conversion model using the Newton method, the color conversion model needs to be differentiable. Therefore, a polynomial or a neural network that is a differentiable function is used as the color conversion model. ing. Recently, Japanese Patent Laid-Open No. 9-1
As shown in Japanese Patent No. 86894, a neural network, which is a more accurate color conversion model, is widely used.

As a method of solving a color conversion model constituted by a neural network without using the above-mentioned CCM,
JP-A-8-102865 and JP-A-8-20497
No. 3 discloses a method. In these methods, the image recording signal CM used to create a color conversion model for obtaining an L * a * b * color signal from the image recording signal CMYK is used.
By making the neural network learn a data set of YK and L * a * b * color signals, L * a * b
*, And a neural network that outputs the image recording signal CMY from the black amount K is created. Using this neural network, the L * a * b * color signal and the black amount K
From the image recording signal CMYK.

[0008]

The above-mentioned JP-A-10-7
A method of solving a color conversion model using CCM proposed in Japanese Patent Application Laid-Open No. 0668 and Japanese Patent Application Laid-Open No. 9-186894 is as follows.
Although it is certainly effective in that the color conversion model is solved using the Newton method, the convergence is fast, and the color conversion coefficient of the direct lookup table can be obtained at high speed. There is a problem that a continuous color conversion model needs to be used, and a differentially discontinuous color conversion model cannot be used.

In the method using a neural network as a color conversion model proposed in Japanese Patent Application Laid-Open No. 9-186894, a highly accurate color conversion coefficient can be obtained using a highly accurate neural network type color conversion model. Although it is certainly effective in that it can be done, the following problems have been found by study of the inventors. For example, a * = b *
In the case where the lightness L * is changed from 100 to 0 for a gray portion where = 0, Japanese Patent Laid-Open No. Hei 9-18649 is used.
FIG. 5 (a) shows the respective values of the component CMYK of the image recording signal obtained by the method proposed in Japanese Patent Publication No. The horizontal axis of FIG. 5A indicates a value obtained by quantizing the lightness L * to 8 bits. A value 0 on the horizontal axis indicates L * = 100, and a value 255 on the horizontal axis indicates L * = 0. ing. The vertical axis indicates the dot area ratio (%) of each color of the image recording signal CMYK.

As shown in FIG. 5A, when the dot area ratio of the black amount K in the gray portion (shadow portion) increases,
A phenomenon is seen in which the dot area ratio of the image recording signal CMY, which is supposed to monotonically increase, reverses and decreases. When such a phenomenon in which the gradation is inverted occurs, there arises a problem of causing a serious defect in image quality such as generation of a false contour in an image, reduction in density, occurrence of moire, and the like. . Since such a phenomenon occurs remarkably when the black amount K is large, there is a problem that it is difficult to set a large value of the black amount K in the related art.

Such a phenomenon is a phenomenon that occurs when a neural network type color conversion model is solved by a Newton method, which is a kind of successive approximation method, because the solution converges in an unexpected local region. In the neural network type color conversion model, there are many unexpected local areas in areas where the amount of black K is large, highlights and shadows, etc., and when a solution is obtained using the Newton method, it converges to a value other than the correct answer. It has been found by the inventor of the present invention that this occurs because an accurate color separation value cannot be obtained.

On the other hand, in order to solve a color conversion model without using CCM, it is generally necessary to obtain a solution by fully searching the color conversion model for an image recording signal CMYK having a finite gradation of 12 bits or less for each color. Is possible, but "High-precision color conversion by flexible UCR ~ (3)
Color Conversion Model by Neural Network ~ ", Kazumasa Murai, Japan Hardcopy '96 Transactions, pp. 2
As shown in FIG. 09-212, a problem arises that the entire search in the four-dimensional space is not realistic because the calculation time is enormous.

As a method for solving a color conversion model constituted by a neural network without using a CCM, an image recording signal CMY used when creating a color conversion model is used.
A method of obtaining an inverse model of a color conversion model by making a neural network learn a data set of K and colorimetric values L * a * b * has been proposed in Japanese Patent Application Laid-Open Nos. 8-102865 and 8-204973. Has been
"High-precision color conversion by flexible UCR-(3) Color conversion model by neural network-", Kazumasa Murai, Japan Hardcopy '96, pp. 146-64. 20
As shown in FIG. 9-212, this method uses the CCM
There is a problem that the color conversion accuracy is lower than the method using CCM, and it is necessary to find the inverse model of the color conversion model by learning the neural network. Therefore, the conversion time is much longer than the method using CCM. The problem arises.

[0014] It is an object of the present invention to provide a color that can quickly obtain a second color signal from a first color signal with accurate color reproducibility based on a color conversion model for obtaining a first color signal from a second color signal. An object of the present invention is to provide an image processing method and apparatus.
Further, a color image processing method capable of rapidly obtaining a second color signal from a first color signal with accurate color reproducibility even if a color conversion model for obtaining a first color signal from a second color signal is not differential and continuous. It is to provide a device.

[0015]

SUMMARY OF THE INVENTION An object of the present invention is to provide a color image processing method for converting a first color signal into a second color signal, and a color conversion method for obtaining a first color signal from the second color signal. Based on the step of obtaining the model and the color conversion model,
Obtaining a second color signal from the first color signal by performing a non-linear optimization method using no derivative of the color conversion model. In addition, the above object is to convert the first color signal into the second
A color image processing device for converting into a color signal,
A color prediction unit having a color conversion model for obtaining a first color signal from the second color signal; and a non-linear optimization method using no derivative of the color conversion model based on the color conversion model, thereby obtaining a first color signal. And a non-linear optimizing means for obtaining a second color signal from the signal.

[0016]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS A color image processing method and apparatus according to an embodiment of the present invention will be described with reference to FIGS. First, a schematic configuration of a color image processing apparatus that determines a color conversion coefficient by the color image processing method according to the present embodiment will be described with reference to FIG. FIG. 1 shows an example of a color image output system including the present color image processing apparatus.

The present color image output system includes an image input device 100, an image processing device 200, and an image forming device 30.
It has 0. The image input device 100 receives color images in various formats from the outside and outputs color image signals. In the present embodiment, the data of each color of R (red), G (green) and B (blue) has a total of 24 bits of 8 bits and 256 gradations.
A color image signal composed of bit RGB data is output.

Specifically, the image input device 100 is 35 m
m color negative film, positive film, or APS
A silver halide photographic film represented by a film
Read as RGB data by sensor or K
ODAK PhotoCD ™ format C
The image data is read from the D-ROM and converted into RGB data, or the image data is read from a digital camera such as DCS1c (trade name) of Canon Inc.
Color image data converted to B data or edited by a user using another computer and stored in a recording medium represented by MO or Zip (storage medium standard) is read from the recording medium and converted to RGB data. Converting, or converting image information transmitted from a device connected on the network into RGB data,
It has a function of transferring the RGB data to the image processing apparatus 200.

The image forming apparatus 300 includes the image processing apparatus 20
0, the image recording signal (YM in this embodiment)
CK data) is formed on paper. FIG. 2 shows a schematic configuration of an example of the image forming apparatus 300 of the present invention. In FIG. 2, a belt-shaped intermediate transfer body 50 includes rollers 5-1, 5-2, 5-3,
5-4 and supported by the heating roll 2 so as to rotate in the direction of the arrow in the figure. The heating roll 3 is arranged to face the heating roll 2. The four photoconductors 1-1, 1-2, 1-3, and 1-4 are provided around the intermediate transfer body 50.
Is arranged. These four photoconductors 1-1 and 1-
2, 1-3, 1-4 are electrostatic latent image forming chargers 15, 1
6, 17, and 18 uniformly charge. The screen generator 390 performs pulse width modulation on the laser light based on the CMYK four-color image recording signals transferred from the image processing device 200. The pulse width-modulated laser light is applied to four photoconductors 1-1 by a laser scanner scanning device 380.
Horizontal scanning is performed on 1-2, 1-3, and 1-4. Thus, an electrostatic latent image is formed on the four photoconductors 1-1, 1-2, 1-3, and 1-4.

Four photosensitive members 1 on which an electrostatic latent image is formed
1, 1-2, 1-3, 1-4, a black developing device 1
1, a yellow developing unit 12, a magenta developing unit 13, and a cyan developing unit 14 form black, yellow, magenta, and cyan toner images, respectively. These toner images are sequentially transferred to transfer units 50-1, 50-2, 50-3, 5
The data is transferred to the intermediate transfer member 50 by 0-4. As a result,
A plurality of color toner images are formed on the intermediate transfer member 50.
Thereafter, a recording sheet having a surface coated with a thermoplastic resin layer is fed from a sheet tray 6 by a sheet feeding device 7.
The recording paper is transferred to a pin roll 9-attached to a winding mechanism 8.
After being heated while being wound around the heating roll 3 by 1, 9-2, it is heated under pressure by the heating roll 2 and the heating roll 3 in a state of being in close contact with the intermediate transfer body 50. Thereby, the toner images of a plurality of colors on the intermediate transfer body 50 penetrate into the thermoplastic resin layer on the recording paper.

The intermediate transfer body 50 and the recording paper, which have been pressurized and heated by the heating rolls 2 and 3, move while being in close contact with each other, and are cooled by the cooling device 4. This causes the toner that has permeated the thermoplastic resin layer to coagulate and solidify,
It produces strong adhesion to recording paper. Thereafter, on the roll 5-1 having a small curvature, the recording paper is separated from the intermediate transfer body 50 together with the toner by the stiffness of the recording paper itself, and is output to the outside. Since the toner image transferred and fixed on the recording paper is integrated with the resin layer on the surface of the recording paper, the surface of the recording paper has a smooth and high gloss.

As the photoconductors 1-1, 1-2, 1-3, and 1-4, various inorganic photoconductors (Se, a-Si, a-Si
C, CdS, etc.) and various organic photoreceptors can be used. As the toner, a known material composed of a thermoplastic binder containing a dye such as yellow, magenta, and cyan can be used. In this embodiment, a polyester toner having a weight average molecular weight of 54000, a softening point of 113 ° C., and an average particle diameter of 7 μm is used. The amount of toner on the recording medium of each color varies depending on the content of the dye in the toner, but is about 0.4 mg / cm ^{2 to} 0.7 mg / c.
As will become m ^2, it is set as the exposure or development conditions. In the present embodiment, each color is 0.65 mg / cm ^2.
It is set to be.

As a recording medium, a commercially available cast-coated paper having a basis weight of 127.9 g / m ² and a 7 μm-thick polyester coated on the surface of an enamel-coated paper (Yonago Kagami Co., Ltd.) is used. . In the present embodiment, the polyester to be coated has a weight average molecular weight of 1230.
A polyester having 0, a number average molecular weight of 3270 and a softening point of 100 ° C. is used.

The intermediate transfer member 50 has a two-layer structure of a base layer and a surface layer. The base layer is
A 70 μm thick polyimide film to which carbon black was added was used, and the volume resistivity was adjusted to 10 ¹⁰ Ωcm by changing the amount of carbon black added. As the base layer, for example, a high heat-resistant sheet having a thickness of 10 to 300 μm can be used. More specifically, polyester, polyethylene terephthalate, polyether sulfone, polyether ketone, polysulfone, polyimide, A polymer sheet of polyimide amide, polyamide, or the like can be used.

In the present embodiment, the intermediate transfer member 50 and the recording sheet are in close contact with each other with the toner image interposed therebetween when the simultaneous transfer and fixing from the intermediate transfer member 50 to the recording sheet are performed. As described above, using a silicone copolymer having a rubber hardness of 40 degrees and a thickness of 50 μm, the photosensitive members 1-1, 1-2, 1-
The volume resistivity of the silicone copolymer is adjusted to 10 ¹⁴ Ωcm in order to transfer a toner image from 3 and 1-4 to the intermediate transfer body 50 without electrostatic image disturbance. Silicon copolymer has a property that its surface is sticky to toner at room temperature and has the property of easily releasing toner that has been melted and fluidized, so that it can efficiently transfer toner to recording media such as recording paper. It is suitable for the surface layer. In addition, as the surface layer, for example, a resin layer having a high releasability having a thickness of 1 to 100 μm can be used, and more specifically, a tetrafluoroethylene-perfluoroalkylvinyl ether copolymer, polytetrafluoroethylene Etc. can be used.

As the heating rolls 2 and 3, metal rolls,
Alternatively, a metal roll having a heat-resistant elastic layer made of silicon rubber or the like can be used. A heat source is disposed inside the heating rolls 2 and 3, and the set temperature of the heat source is determined by the thermal melting characteristics of the toner and the thermoplastic resin layer on the surface of the recording paper. In this embodiment, since the softening point of the toner is higher than the softening point of the resin layer, the temperature of the heat source is set so that the setting temperature of the heating roll 2> the setting temperature of the heating roll 3. Specifically, the set temperature of the heat source of the heating roll 2 is set to 150 ° C., and the set temperature of the heat source of the heating roll 3 is set to 1
It is set at 20 ° C. In the present embodiment, the pressure between the heating roll 2 and the heating roll 3 at the time of transfer and fixing is 5 kgf /
Although it is set so that the cm ^2, the pressure is not limited thereto, may be in the range for example of ^{1kgf / cm 2 ~10kgf / cm 2} . Further, in the present embodiment, the outer diameter of the heating roll 2 and the heating roll 3 is set to 50 mm, and the rotation speed of each of the heating rolls 2 and 3 is set to 2
It is set to be 40 mm / sec. Further, in the present embodiment, the temperature of the surface of the recording medium in contact with the intermediate transfer member 50 when the recording medium such as the recording paper is separated from the intermediate transfer member 50 is adjusted by adjusting the air volume of the cooling device 4.
The temperature is set to 0 ° C.

In the present embodiment, a tandem engine electrophotographic color printer is used as the image forming apparatus 300. However, the present invention is not limited to this, and a single engine type or a heat resistant belt without using an intermediate transfer member is used. A printer of the type in which a plurality of color toner images formed on a belt photoreceptor are directly transferred and fixed on recording paper using a photoreceptor may be used. Further, the image forming apparatus 300 is not limited to an electrophotographic color printer, but may be a color image forming apparatus such as a printing, ink jet type, thermal transfer type, and silver halide photographic type.

The image processing apparatus 200 includes a first color conversion unit 21
0 and a second color conversion unit 220. The RGB data input from the image input device 100 is converted by the first color conversion unit 210 into CIE L *, which is one of the uniform color spaces.
It is converted into data in the a * b * color space. First color conversion unit 2
The L * a * b * data (first color signal) converted by 10 is converted by the second color conversion unit 220 into the image forming apparatus 3.
The image recording signal in the color space of 00, in this embodiment, Y
(Yellow), M (magenta), C (cyan) and K
The image data is converted into four-color (black) data (second color signal) and transferred to the image forming apparatus 300. Image input device 1
In this embodiment, the most common RGB data will be described as an example of an input color signal from 00, but other color spaces such as a CMYK color space used in printing and a YCC color space used in PhotoCD are used. Data may be used.

In the present embodiment, the first color conversion unit 210 converts an input color signal into a signal in an L * a * b * color space. However, the present invention is not limited to this.
The signal may be converted into a signal in another color space that does not depend on a device such as a Z color space or an L * u * v * color space.
It is more preferable that the first color conversion unit 210 converts the signal into a signal of a uniform color space such as L * u * v *. As the first color conversion unit 210, a matrix operation type color conversion circuit, a direct look-up table type color conversion circuit, or a neural network type color conversion circuit widely used as a color conversion circuit may be used. This is possible, and a color conversion circuit of a neural network type is used in the present embodiment.

The first color converter 210 can be obtained by the following method. First, a colorimetric value (L * a * b *) of a color image input to the image input device 100 is measured with a commercially available colorimeter, and a colorimetric value (L * a * b) of the input color image is measured.
*) RGB data corresponding to the input colorimetric value (L
* A * b *) to model the color conversion characteristics of output data (RGB). Higher-order polynomials and neural networks are generally used for such color conversion models,
In the present embodiment, the neural network is made to learn the data set of the combination of the RGB data and the L * a * b * data, and the conversion characteristics of the image input device 100 are modeled. The obtained neural network is the first
The color conversion unit 210.

Next, the second color converter 220, which is a main part of the present invention, will be described. In the present embodiment, the second color conversion unit 220 is configured by a three-input four-output direct lookup table. The direct lookup table is configured to calculate an image recording signal CMYK by performing an interpolation operation by cubic interpolation using a value obtained by dividing each axis of the input signal L * a * b * by 16 as an input address.
In the present embodiment, the cubic interpolation method is used as the interpolation method of the direct lookup table, but the interpolation method is not limited to this, and other interpolation methods such as a triangular prism interpolation method and a tetrahedral interpolation method are used. May be applied. Further, in the present embodiment, the second color conversion unit 220 is configured by a direct look-up table. However, the present invention is not limited to this configuration, and other configurations may be used as long as color conversion of three inputs and four outputs can be performed. Good.

In the present embodiment, the second color converter 22
The process of determining the direct lookup table constituting 0 will be described with reference to FIG. First, in step S1, a plurality of color patches for an arbitrary combination of image recording signals CMYK are printed out by the image forming apparatus 300, and the colorimetric value L * a of each printed color patch is printed using a colorimeter. * B * is measured. In the present embodiment, the combination of the image recording signals CMYK is
6.times.6.times.6.times.6 = 12 where the halftone dot area ratio of each color is 20%.
The image forming apparatus 300 printed out a combination of 96 patches. As a colorimeter, X-Rite 938 (trade name) which is a colorimeter of X-Rite was used.
L * a * b with the measurement conditions of the colorimeter set to D50 and a 2 degree field of view
* Was measured. Although the number of color patches used for measurement may be arbitrary, it is desirable to increase the number as much as possible in order to improve the accuracy of the color conversion model. In the present embodiment, the L * a * b * color system, which is a uniform color space, is used as the color system used for measurement. However, another color system such as the XYZ color system may be used. It may be. In the case where the color difference is evaluated when solving the color conversion model, a uniform color space is preferable.

Next, in step S2, step S
The neural network learns the data set of the plurality of image recording signal CMYK data obtained in step 1 and the corresponding L * a * b * data as teacher data. CM
The relationship between YK and L * a * b * can be represented by the following function. (L *, a *, b *) = F (C, M, Y, K) (1) When this equation (1) is decomposed into respective color components, the following is obtained. L * = FL (C, M, Y, K) (2) a * = Fa (C, M, Y, K) (3) b * = Fb (C, M, Y, K) (4) In the present embodiment, the neural network is
Document "High-precision color conversion by flexible UCR-high-precision printer model by neural network-", Kazumasa Murai, Japan Hard Copy '94 Transactions, pp. 1-64. The learning was performed by a back-provacation method using a neural network having four layers and the number of intermediate layers of 10 cells shown in 181-184. In the present embodiment, the color conversion model of the neural network is used in step S2, but another polynomial model or a conversion table type color conversion model may be used. As will be described later, according to the present invention, even a color conversion model that is differentially discontinuous can be solved, so that the range of selection of a color conversion model is wider than before.

Next, in step S3, the maximum black amount Kmax and the minimum black amount Kmin for reproducing the input address value L * a * b * of the direct look-up table are determined by the following method based on the neural network. Calculate and calculate the optimal black amount K from Kmax and Kmin.
To determine. Here, normally, the inverse function of the function F shown in Expression (1) is not obtained. However, if the value of L * a * b * is given and one variable in CMYK is appropriately determined, the remaining three variables can be obtained from equation (1). For example, the value of L * a * b * is given to equation (1),
, The value of CMY can be determined.

Here, a method of determining the value of CMY will be considered. In the prior art such as CCM, CMY when L * a * b * and K are given to equation (1)
Was calculated by a successive calculation method such as Newton's method, which is one of the solutions to nonlinear equations. For example, in the Newton method, since a solution is searched for using the derivative of the equation (1), the solution converges to an unexpected local region as described above, and an accurate color separation value cannot be obtained. Problems arise. Therefore, in the present invention, equation (1) is solved by applying a nonlinear optimization method that does not use the derivative of equation (1). Here, assuming that the color to be reproduced is L * a * b * and the amount of black to be applied is K, the color difference ΔE between the color to be reproduced and the color based on the image recording signal CMY and the black amount K is the value of the image recording signal CMY. It is defined by the following equation as a function. ΔE (C, M, Y) = ((L * −FL (C, M, Y,
K)) ² + (a * -Fa (C, M, Y, K)) ² + (b
* -Fb (C, M, Y, K)) ² ) ^1/2 ...
(5) Solving equation (1), which is a non-linear equation, means the same as finding the value of CMY where the color difference ΔE is zero. CM that minimizes the objective function ΔE using the color difference ΔE as the objective function
It can be reconsidered as a nonlinear optimization problem of finding Y. Therefore, solving Equation (5) using a non-linear optimization method that does not use derivatives, such as the simplex method,
That is, CMY can be obtained. Therefore,
Even if the color conversion model is discontinuous, the value of CMY can be obtained.

Although the simplex method is known, for example, "Nonlinear Programming", Hiroshi Konno, Nikka Giren Publishing, pp. The algorithm is introduced in 284-287. The simplex method is a technique suitable for optimizing such a multivariable function, and can obtain an optimum value at high speed. The simplex method does not require derivatives of multivariable functions unlike the Newton method, so that the solution does not converge to an unexpected local region of the neural network, and an accurate solution can be obtained. Since it is possible to optimize a differentially discontinuous function, there is an advantage that the range of selection of a color conversion model is widened. In the present embodiment, a simplex method capable of rapidly optimizing a multivariable function is applied as a nonlinear optimization method that does not use the derivative of the color conversion model. Other non-linear optimization methods that do not use the derivative of may be applied.

First, while varying the value of K on the right side of equation (5), it is determined whether or not the CMY value obtained satisfies the condition shown in the following equation. 0 ≦ C, M, Y ≦ 100 (%) (6) Here, the largest value of K satisfying the condition is the maximum black amount Km.
ax, and the smallest value of K is the minimum black amount Kmin. Since the value of the black amount K is usually quantized to about 8 bits, even if the value of K is fully searched from 0 to 100%, 256
By optimizing equation (5), Kmax and Kmin satisfying equation (6) can be obtained. Also,
Expression (5) by a binary search algorithm without performing a full search
Kmax and Kmin satisfying the expression (6) can be efficiently obtained by varying the value of K from 0 to 100%.

Here, without finding the minimum black amount Kmin, it is assumed that the minimum black amount is the case where blackening is not performed, and Kmin =
It may be set to 0. In the present embodiment, Kmin = 0 is set without finding the minimum black amount Kmin. In the present embodiment, the optimum black amount K is determined by the following equation. K = α · β · Kmax + (1−α · β) · Kmin (7) where the variables α and β are functions of lightness L * and chroma C *, respectively, and are shown in FIG. Has characteristics. FIG.
The function shown in (a) represents the inking rate α with respect to the lightness L *, and is represented such that the density increases in the X-axis increasing direction, that is, the value of L * decreases. FIG.
The function shown in (b) represents the inking ratio β for the saturation C *. Note that C * = (a * ² + b * ² ) ^1/2 (8).

As shown in FIG. 4A, in the function of the inking rate α relating to the lightness L *, black K is mixed in a portion having a medium density or less, that is, in a bright portion where the brightness is high, and the granularity of the image is reduced. In order to prevent the image quality from worsening and to prevent the occurrence of moiré, the inking ratio α is set to 0% so that the black ink is not applied to a bright portion where the value of L * is 50 or more. Also, the inking rate α at the point where the value of L * is 50 is set to 0%, the inking rate α at the point where the value of L * is 0 is set to 100%, and the point between the two points is convex downward. Is set to follow the quadratic function of Therefore, since black does not appear on a bright portion having a medium density or less, the graininess is excellent and moire does not occur. Further, the value of L * is 0 to 50.
In the part between, the inking rate α according to the quadratic function convex downward
Is increased, the difference at the beginning of black insertion can be reduced, and a false contour or the like due to mixing of black can be prevented.

In the present embodiment, the inking ratio α is 0 in brighter portions than the point where the value of L * is 50.
%, But the value of L * at the reference point is not limited to this, and it is desirable that the value of L * be set in the range of 70 to 30. Further, in the present embodiment, the inking ratio α at the maximum density point (the point where the value of L * is 0) is set to 100%, but the value of the inking ratio α is not limited to 100%. It is desirable to set in the range of 100%. Further, in the present embodiment, a function including a quadratic function is used as a function of the inking rate α related to the lightness L *, but the present invention is not limited to this, and a function including an arbitrary function such as a linear function is used. May be used, and a free function based on a lookup table obtained by conducting an experiment may be used.

As shown in FIG. 4B, in the function of the inking ratio β relating to the saturation C *, black is mixed in a high saturation part, the saturation is reduced, and the graininess is deteriorated. In order to prevent this, the inking ratio β is set to 0 so that the ink K will not be applied to the vivid portions beyond the value of C * with reference to 40.
% Is set. Further, the inking ratio β at the point where the value of C * is 40 is set to 0%, the inking ratio β at the point where the value of C * is 0 is set to 100%, and a linear function between the two points is set. It is set to follow. Therefore, the vivid areas of the set saturation or higher will not get any black ink,
Saturation does not decrease in a high saturation portion, and color reproduction with excellent granularity can be performed.

In this embodiment, the value of C * is 40
Although the inking ratio β of the bright portion is set to 0%, the value of the reference C * is not limited to this, and is 20 to 20%.
It is desirable to set in the range of 60%. Further, in the present embodiment, since the basic inking ratio is controlled by the inking function α relating to the lightness L *, the inking ratio in the gray portion (point where the value of C * is 0) is set to 100. %, But the inking ratio β at the point where the value of C * is 0 is not limited to 100% and may be set to another value. Also, in the present embodiment, the function of the inking ratio β for the saturation C * is 1
Although the function is set to a function including a quadratic function, the type of the function is not limited to this. The function can be set to a function including an arbitrary function such as a quadratic function. A free function using an up-table may be set. In equation (7), the change in the inking amount β with respect to the saturation C * is not considered, that is, β is set to 100%
Alternatively, the inking amount K may be controlled in accordance with the inking function α of only the lightness L *.

Next, in step S4, the objective function ΔE of equation (5) is minimized from the input address value L * a * b * of the direct lookup table and the optimum black amount K obtained by equation (7). The value of CMY to be converted is determined using the simplex method. Next, in step S5, the CMY values obtained in step S4 and the optimal black amount K
Set to the grid point of the direct lookup table corresponding to * a * b *. Thereby, the second color conversion unit 22
It is possible to determine the direct lookup table that constitutes 0.

In the present embodiment, the color space of the image forming apparatus 300 is a YMCK color space including black, but may be another color space such as a YMC color space or RGB color space not including black K. In this case, assuming that the black amount K is zero, the direct lookup table can be determined in the same manner as described above. As an example showing the effect of the present embodiment, for example, gray (shadow portion) of a * = b * = 0 in which the lightness L * changes from 100 to 0 is shown in FIG. 4 using the color image processing method of the present invention. FIG. 5B shows the values of each CMYK when color separation is performed using the inking function. The horizontal axis of FIG. 5B indicates a value obtained by quantizing the lightness L * to 8 bits, where 0 indicates a value of L * of 100 and 25
5 indicates the value 0 of L *. The vertical axis indicates the dot area ratio (%) of each color of the image recording signal CMYK.

In the prior art shown in FIG. 5 (a), as described above, when the black amount K in the shadow portion increases, the phenomenon that the image recording signal CMY, which is supposed to monotonously increase, is inverted and the value decreases. However, according to the present embodiment, as shown in FIG. 5B, the image recording signal CMY monotonously increases as it should be without inverting the gradation. As described above, according to the present invention, it is possible to prevent the occurrence of a false contour, the decrease in density, and the occurrence of moire as seen in the related art, and to obtain good color reproduction. Further, even when the black amount K is considerably large, a phenomenon unlike the related art does not occur, so that the value of the black amount K can be set large.

As described above, a color conversion model for obtaining tristimulus values such as L * a * b * of a color system from a device color signal such as a CMYK signal is obtained by learning a neural network. By solving the model using the simplex method, which is one of the nonlinear optimization methods that do not use the derivative of the color conversion model, L * a * b *
When obtaining the image recording signal CMYK from the color signal and the black amount K, the neural network type color conversion model can be solved more accurately and faster than before, and accurate color reproduction can be obtained. Further, even when the value of the black amount K is set to be large, the inversion phenomenon of the gradation does not occur.
Good color reproduction without false contours and density reduction can be performed. In addition, the color conversion model can be solved without being constituted by a differential and continuous model such as a neural network, and the options of the color conversion model can be expanded.

Next, another example of the image processing apparatus 200 which is a main part of the present invention will be described with reference to FIG. The second color conversion unit 220 of the image processing apparatus 200 shown in FIG. 6 is not configured with a direct look-up table, and has a configuration in which a corresponding CMYK signal is appropriately obtained from each input L * a * b * signal. ing. In addition, each input L * a * b
The basic concept for obtaining the corresponding CMYK signal from the * signal is the same as in the above example in which the second color conversion unit 220 is configured by a direct look-up table. First, first
The L * a * b * data obtained from the color conversion unit 210 is input to the maximum black amount determination unit 221 and the minimum black amount determination unit 222. Maximum black amount determining unit 221 and minimum black amount determining unit 222
Has a color prediction unit, a nonlinear optimization unit, and a binary search unit. The color prediction unit has a color conversion model that predicts colors L * a * b * reproduced from the image recording signal CMYK of the image forming apparatus 300, and is configured by a neural network. This neural network
Similar to the neural network when creating the direct lookup table of the second color conversion unit 220, it is created by learning a plurality of pairs of CMYK and L * a * b * of the target printer in advance as teacher data. Is done. Here, the relationship between CMYK and L * a * b * can be expressed by Expression (1), Expression (2), Expression (3), and Expression (4).

The non-linear optimizing unit includes the second color converting unit 22
0, the L * a * b * data obtained from the first color conversion unit 210 and a predetermined value are determined using the simplex method, which is a nonlinear optimization method using no derivative, as in the case of creating the direct lookup table of 0. Based on the black amount K and the equation (5), the CMY for minimizing the objective function ΔE is efficiently calculated. In the binary search unit, the value of CMY obtained by the non-linear optimization unit satisfies the condition of expression (6) while shaking the value of K on the right side of equation (5) used in the non-linear optimization unit by the binary search algorithm. It is determined whether or not. The largest K value that satisfies this condition is the maximum black amount Kmax, and the smallest K value is the minimum black amount Kmin. The maximum black amount determination unit 221 and the minimum black amount determination unit 222 are provided with a maximum black amount Kma.
x and the minimum black amount Kmin are output to the black amount determination unit 224.

Also, a obtained from the first color conversion unit 210
* And b * are input to the saturation determination unit 223. The saturation determining unit 223 calculates the a * and b * input by the equation (8).
, And outputs it to the black amount determination unit 224. In the black amount determining unit 224, the lightness signal L * obtained from the first color conversion unit 210, the saturation signal C * obtained from the saturation determining unit 223, and the maximum black amount Kmax obtained from the maximum black amount determining unit 221. Based on the minimum black amount Kmin obtained from the minimum black amount determining unit 222 and the minimum black amount determining unit 222, the optimum black amount K is determined by the blackening function shown in Expression (7), and the CMY determining unit 226
To the CMYK output unit 227. The user can change the inking function using the parameter input unit 225.

The CMY determination section 226 has a color prediction section and a non-linear optimization section similar to the above. In the CMY determining unit 226, the three-color signal L * a * obtained from the first color converting unit 210
The image recording signal CMY of the image forming apparatus 300 is determined by the non-linear optimizing unit from b * and the black amount K obtained from the black amount determining unit 224, and is sent to the CMYK output unit 227.
The CMYK output unit 227 outputs to the image forming apparatus 300 an image recording signal CMYK in which the three colors of CMY obtained from the CMY determining unit 226 and the black amount K obtained from the black amount determining unit 224 are combined.

As described above, the second color conversion section 220 is not constituted by a direct look-up table, but is
Is obtained, it is possible to further prevent the occurrence of a pseudo contour due to an interpolation error which becomes a problem in the direct look-up table, and to realize a very high-precision color conversion. In addition, since the simplex method that can obtain a solution at high speed is used in the nonlinear optimization method, the color conversion is performed at a higher speed than the conventional case when the configuration of the present embodiment is realized by software. be able to.

In this embodiment, the image forming apparatus 3
Although an example in which a signal in the YMCK color space is output to 00 has been described, the present invention can be applied to a case in which a signal in another color space such as a YMC color space or an RGB color space that does not include black K is output. Can be realized by the same processing as described above, provided that the black amount K is set to zero.

[0053]

As described above, according to the present invention, based on the color conversion model for obtaining the first color signal from the second color signal, the second color signal from the first color signal can be accurately reproduced at high speed. You can ask. Further, even if the color conversion model for obtaining the first color signal from the second color signal is not differentially continuous, the second color signal can be obtained from the first color signal at high speed with accurate color reproduction.

[Brief description of the drawings]

FIG. 1 is a diagram showing an example of a color image output system having a color image processing device according to an embodiment of the present invention.

FIG. 2 is a diagram illustrating an example of a color image forming apparatus according to an embodiment of the present invention.

FIG. 3 is a diagram illustrating an image processing method according to an embodiment of the present invention.

FIG. 4 is a diagram showing an example of an inking function according to an embodiment of the present invention.

FIG. 5 is a diagram illustrating an example of a conversion result of a color signal according to a conventional example and an embodiment of the present invention.

FIG. 6 is a diagram illustrating another example of the second color conversion unit of the image processing device according to the embodiment of the present invention;

[Explanation of symbols]

 1-1 to 1-4 Photoconductor 2 Heating roll 3 Pressure roll 4 Cooling device 5-1 to 5-4 Roller 6 Paper tray 7 Paper feeder 8 Winding mechanism 9-1, 9-2 Pin roll 11 Black developing device DESCRIPTION OF SYMBOLS 12 Yellow developing device 13 Magenta developing device 14 Cyan developing device 15-18 Charger for forming an electrostatic latent image 50 Intermediate transfer body 50-1-50-4 Transfer device 100 Image input device 200 Image processing device 210 First color conversion unit 220 Second color conversion unit 221 Maximum black amount determining unit 222 Minimum black amount determining unit 223 Saturation determining unit 224 Black amount determining unit 225 Parameter input unit 226 CMY determining unit 227 CMYK output unit 300 Image forming device 380 Laser scanner scanning device 390 Screen generator

Continuing on the front page (72) Inventor Kazuhiro Iwaoka 430 Nakai-cho Sakai-Kamigami-gun, Kanagawa Prefecture Green Tech Nakai Inside Fuji Xerox Co., Ltd. (72) Inventor Masahiro Takamatsu 430 Nakai-machi Sakai, Ashigara-Kan Kanagawa Prefecture Green Tech Nakakai Fuji Xerox Co., Ltd. F-term (Reference) 5B057 AA11 BA28 CA01 CA08 CA12 CA16 CB01 CB07 CB12 CC01 CD11 CE18 CH07 5C066 AA05 BA20 CA01 CA08 EE04 GA01 HA03 KE01 KE09 KF05 KP05 5C077 MP08 PP15 PP31 PP32 PP33 PP36 PP38 PP41 PQ15 PQ23B03 H03 H03 H03 H03 H03 MA13 PA02 PA03

Claims (12)

[Claims] 1. A color image processing method for converting a first color signal into a second color signal, comprising: obtaining a color conversion model for obtaining the first color signal from the second color signal; Obtaining a second color signal from the first color signal by performing a non-linear optimization method without using a derivative of the color conversion model based on the color conversion model. . 2. A color image processing method according to claim 1, wherein said nonlinear optimization method is a simplex method. 3. The color image processing method according to claim 1, wherein the color conversion model is a neural network. 4. The color image processing method according to claim 1, wherein the first color signal is a signal of tristimulus values of a color system, and the second color signal is a color image input device. Alternatively, the color image processing method is a device color signal of a color image forming apparatus. 5. The color image processing method according to claim 4, wherein the tristimulus value signal of the color system is a signal of a uniform color space. 6. The color image processing method according to claim 4, wherein the second color signal is a device color signal of four colors including black of a color image output device. . 7. A color image processing apparatus for converting a first color signal into a second color signal, comprising: a color prediction model having a color conversion model for obtaining the first color signal from the second color signal. And a non-linear optimization means for obtaining the second color signal from the first color signal by performing a non-linear optimization method using no derivative of the color conversion model based on the color conversion model. Characteristic color image processing device. 8. A color image processing apparatus according to claim 7, wherein said non-linear optimization means uses a simplex method as said non-linear optimization method. 9. The color image processing apparatus according to claim 7, wherein said color conversion model is a neural network. 10. The color image processing apparatus according to claim 7, wherein said first color signal is a signal of a tristimulus value of a color system, and said second color signal is a color image input device. Or a device color signal of a color image forming apparatus. 11. The color image processing apparatus according to claim 10, wherein the signal of the tristimulus value of the color system is a signal of a uniform color space. 12. A color image processing apparatus according to claim 10, wherein said second color signal is a device color signal of four colors including black of a color image output device. .