JP2002010095A - Color image processing method and color image processor - Google Patents

Abstract

PROBLEM TO BE SOLVED: To form a highly precise color correction table with easy and few sampled color charts, and realize highly precise color correction to colors except the sampled color charts. SOLUTION: On the basis of an estimation function derived from between input color chart data and corresponding read color chart data, symbolic recombination processing is performed to each candidate in a candidate group of previously prepared plural color correction matrixes, and a color correction matrix having an optimum estimation value is presumed. By using the color correction matrix, conversion is performed from digital image data to image data after color correction which are to be outputted with an output machine.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a color processing technique for digital color image data, and more particularly to a color image processing method and a color conversion method for reproducing the same color as an original on a color copying machine, a color printer, a color monitor and the like. It is about.

[0002]

2. Description of the Related Art In recent years, copying machines and printers have been used on a daily basis, and recently, color copying machines for copying color documents,
2. Description of the Related Art Color printers that print color documents from computers have become widespread.

[0003] Conventionally, when printing or copying a color document, it has been pointed out that the color of the document or image output by the device is different and sufficient color reproducibility cannot be ensured.
This is because the color of the printed matter differs depending on the type of the output device because the pigment is a toner, a dye, or a pigment. In addition, while the RGB color data handled by the computer is colored by additive color mixing, the printer prints by subtractive color mixing,
This is because color mixing is often present in the dye itself due to unnecessary absorption of the dye.

Various measures have been taken to solve the problem of color reproducibility between devices for color images, and the color reproducibility of devices has been gradually improved.

For example, in an electrophotographic color copying machine, RGB color data obtained by a color scanner is
After the color converter corrects the color to color data that matches the characteristics of the printing color material, the data is input to a color engine and printed. The color converter is performed by a matrix calculator,
Some do with a multi-dimensional lookup table.

Hereinafter, as a conventional example, a case in which color correction is performed by a color correction unit constituted by a matrix calculator will be described with reference to the block diagram of FIG.

[0007] When the color correction matrix is not estimated, the input data (usually RGB) is converted by the data conversion means 5 according to the format of the conversion data into the color correction matrix as shown in (Equation 1). The signals are converted into the separation signals Dr, Db, and Dg, and are converted into CMY ink signals by the color correcting means 11 according to (Equation 2). This is because the color corresponding to the R data cannot usually be reproduced with a single ink, but is reproduced by mixing CMY inks. The same applies to G data and B data.

[0008]

(Equation 1)

[0009]

(Equation 2)

Here, Mij is a 3 × 3 matrix, and the least-squares estimating means 190 depends on the presence or absence of the instruction signal 2.
Is obtained as follows.

First, the learning color chart generating means 3 creates an input color chart data set Pi = (C, M, Y) i covering a wide input color gamut according to an instruction signal 2 indicating an instruction for estimating a color correction matrix. Then, a color patch image including the color patch data is output from the printer. Next, the output color chart is measured with a spectrophotometer, and tristimulus values Ti = (X, Y,
Z) i is obtained, and the combination of Pi and Ti is stored in the learning color chart data set 4. FIG. 20 shows an example of a correspondence table between Yi of the input color patch data set and Ti of the read color patch data set. Then, the data conversion means 5 converts the tristimulus values from the learning color chart data set 4 into color separation signals D through RGB signals.
i = (Dr, Dg, Db) i. The data set Di of tristimulus values obtained from the read color patch data set is converted by a 3 × 3 matrix M, and a matrix M # best which is the closest conversion to the input color patch data set Pi is obtained by the least square estimating means 190.
Asks. In this case, the matrix coefficient M # ij is (Equation 2)
Of the converted color chart data set converted from the spectrometer output Ti = (X, Y, Z) i by Poi = (Co, Mo, Yo) i and the input color chart data set value Pi = (C, M, Y) ) Determined by the least squares method so that the sum of squares of the error with i is minimized.

The condition of the least squares method is that each matrix coefficient M #
The matrix coefficient M # ij that minimizes the error is obtained using ij as a variable, and as shown in the following (Equation 3), the partial differential coefficient using the value of the matrix coefficient M # ij as a variable 0
Can be obtained. Here, nnn represents the number of data sets.

[0013]

(Equation 3)

The obtained matrix coefficient M # ij is stored in the color correction coefficient database 10 and the color correction means 11 converts the information of the color correction unit shown in (Equation 1) and (Equation 2) by using the information of this 10. Characteristics are required.

As described above, in the conventional example, the linear conversion matrix is optimized by the least square method so as to minimize the error with respect to the entire combination of color patches for calculating the conversion parameters. In the case of deviation from the generalized linear matrix relationship, there is a problem that a partial error occurs in some color regions. As this improvement, a case of performing a nonlinear masking approximation having a quadratic term as in (Equation 4) is often used. In this case, although the color correction accuracy is improved as compared with the linear masking approximation, there is a problem that when the number of samples is small, a large color shift is likely to occur when correcting a color that is slightly shifted from the sample.

[0016]

(Equation 4)

In order to reduce the calculation load of the higher-order color correction matrix M # ij, the color correction means 11 converts the higher-order matrix calculation into a three-dimensional table instead of the color correction coefficient database 10. A technique for speeding up the color correction processing using a look-up table (hereinafter referred to as “3D-LUT”) is also used. Although the detailed contents and operation of the color corrector using the 3D-LUT are described in JP-A-7-99586 and JP-A-7-99587, input image data (RGB data, etc.) Separate into lower bits and select multiple table data close to the input from the 3D-LUT based on the upper bits. Then, the plurality of table data selected by the lower bits are interpolated, and a color correction output CMY value corresponding to the input RGB value is output. Using the obtained coefficient AMij (Equation 4)
A three-dimensional color conversion table is created by converting the input / output relationship of (Equation 5) into a three-dimensional table. Conventional example 2 improves the drawback of linear correction by increasing the order of the matrix relationship,
In order to further reduce the operation load of the high-order matrix operation, the high-order matrix operation is converted into a multidimensional table. There is also a method in which 11 uses a learning color chart data set of 4 instead of the 10 color correction coefficient databases to perform color correction. At this time, the print values for all the inputs are predicted and calculated from the colorimetric values of the printed color patches by triangular prism interpolation, and the target print color is determined by searching a correspondence table with the predicted colorimetric values for all the inputs. The color correction value to be given is obtained, and is reported in the 1994 Color Forum Annual Conference Proceedings (pp. 55-58). Japanese Patent Laid-Open No. Hei 6-221 discloses a method of creating a color correction table for outputting an arbitrary print color by an interpolation operation in a triangular pyramid surrounded by four adjacent color chips in a color space.
No. 24, which is incorporated herein by reference. At this time, the color correction table values can be directly obtained even in the case of color correction that is difficult to model, but it is necessary to create a large number of data set correspondence tables.

[0018]

As described above, the prior art of the color correction method has a certain effect on the color reproducibility, but each of them has a problem. The present invention has been made in view of such a point, and can realize high-accuracy color correction easily and with a small number of sample color patches and realize high-precision color correction for colors other than the sampled color patches. It is an object to provide a color image processing method and a color conversion method.

[0019]

In order to solve the above-mentioned problems, a first color image processing method and a color image processing apparatus according to the present invention provide an evaluation derived between input color patch data and corresponding read color patch data. Based on the function, a color correction matrix having an optimum evaluation value is estimated by a genetic algorithm that performs a symbolic rearrangement process on each candidate in a plurality of color correction matrix candidate groups prepared in advance.

In order to solve the above-mentioned problems, a second color image processing method and a color image processing apparatus according to the present invention provide a color correction matrix that optimally realizes correspondence between an input color patch data set and a corresponding read color patch data set. When estimating, the method of using the gradient information of the evaluation value is combined with the genetic algorithm used in the first embodiment of the present invention to improve the color correction accuracy and reduce the estimation speed. is there.

In order to solve the above-mentioned problems, a third color image processing method and a color image processing apparatus according to the present invention provide a color image processing apparatus for optimally realizing a correspondence between an input color patch data set and a corresponding read color patch data set. When the correction matrix is estimated, the color correction accuracy is improved by once estimating using the genetic algorithm used in the first embodiment of the present invention and then using the neural network method again. is there.

According to a fourth aspect of the present invention, there is provided a color image processing method and a color image processing apparatus according to the present invention, in which the entire area of an input color range is subdivided, and statistical analysis of input color patch data sampled from each area is performed. By selecting a plurality of input color patch data sets representing each area and the corresponding read color patch data based on the distribution, it is possible to estimate a color correction matrix that realizes highly accurate color correction conversion with a small number of samples. Is what you do.

According to a fifth aspect of the present invention, there is provided a color image processing method and a color image processing apparatus for reading read color patch data corresponding to a plurality of input color patch data sets visually important in an input color range. By selecting and interpolating between the selected input color patch data and the read color patch data, preparing an input color patch data set sufficient for color correction matrix estimation and a corresponding read color patch data set, The color correction matrix can be easily created by the number.

To solve the above problem, a sixth color image processing method and a sixth color image processing apparatus according to the present invention are obtained by fusing the fourth and fifth color image processing apparatuses according to the present invention. Based on the distribution, select multiple input color patch data sets representing the entire input color range and read color patch data, and interpolate between the selected input color patch data and between read color patch data to correct the color By preparing an input color patch data set sufficient for matrix estimation and a corresponding read color patch data set, the accuracy of estimating the color correction matrix with a small number of samples is further improved.

[0025]

Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram of a color image processing apparatus according to a first embodiment of the present invention.
Is a configuration diagram of a color image processing device according to a second embodiment of the present invention, FIG. 8 is a configuration diagram of a color image processing device according to a third embodiment of the present invention, and FIG. FIG. 1 is a configuration diagram of a color image processing apparatus according to a fourth embodiment.
6 is a configuration diagram of a color image processing device according to a fifth embodiment of the present invention, and FIG. 19 is a configuration diagram of a color image processing device according to a sixth embodiment of the present invention. FIG. 13 shows a configuration diagram of a representative learning color chart selecting means in the color image processing apparatus according to the fourth embodiment of the present invention.

FIG. 2 is a flowchart of a process performed by the color image processing apparatus according to the first embodiment of the present invention.
FIG. 7 is a flowchart of the process of the color image processing apparatus according to the second embodiment of the present invention, and FIG.
FIG. 12 is a flowchart of the processing of the color image processing apparatus according to the fourth embodiment, FIG. 12 is a flowchart of the entire processing of the color image processing apparatus according to the fourth embodiment of the present invention, and FIG. FIG. 20 is a flowchart showing the overall processing of the color image processing apparatus according to the fifth embodiment.
Shows a flowchart of the overall processing of the color image processing apparatus according to the sixth embodiment of the present invention. FIG. 14 is a flowchart of the process in the representative learning color chart selecting means in the color image processing apparatus according to the fourth embodiment of the present invention. In the drawings of the configuration, the same portions are denoted by the same reference numerals.

(First Embodiment) First, the first embodiment of the present invention will be described.
A color image processing apparatus according to the embodiment will be described. In FIG. 1, reference numeral 1 denotes color input data input to the apparatus, and any color system can be used as a value, but here, sRGB system is used. 2
Is an instruction signal for the present apparatus to estimate a color correction matrix. Reference numeral 3 denotes learning color chart generation means for generating an input color chart data set for learning by equalizing the entire area of the input color range according to 2, and outputs the generated data by a target printer, and outputs the data. A color information data set (read color chart data set) at the time of output is obtained by measuring the sample with a spectrophotometer or the like, and an input color chart data set is obtained as shown in FIG. Are stored as a set.
Reference numeral 5 denotes data conversion means for converting the read color patch data in the learning color patch data held in 4 into a color separation signal D (Dr, Dg, Db).

6 is a color correction matrix to be estimated.
Initial candidate setting means for setting an initial candidate matrix set GM = [M [k]] (k = 1,..., N) of M by a predetermined procedure.
Is an evaluation value deriving unit that calculates an evaluation value E [k] of each M # k in the candidate matrix set GM according to a preset evaluation function. 8 is an evaluation value E [obtained by the evaluation value deriving unit 7. k] is a fitness calculation means for calculating the fitness f [k] of each M [k] to the problem based on the fitness obtained by the fitness calculation means 8. [k] This is a rearrangement operation means for executing a rearrangement operation process such as selective selection of a matrix and crossover, mutation and the like. Reference numeral 10 denotes a color correction coefficient database for holding the optimum M # best when the estimation of the color correction matrix by the genetic algorithm satisfies a preset convergence condition. Using the optimal color correction matrix M # best in the database and the color separation signal D (Dr, Dg, Db) obtained by converting the input color data 1 by the data conversion means 5, the input data after color correction is obtained. This is color correction means for estimating color data.

The operation of the color image processing apparatus according to the first embodiment configured as described above will be described. First, it is determined whether or not there is an instruction signal 2 for instructing color correction estimation for the input color data 1. If not, the color of the input data is corrected using the color correction matrix M # best in the current color correction coefficient database.
At this time, one input data is converted by the data conversion means 5 into a color separation signal D (Dr, DG, Db). (Equation 1) is used for this conversion. In this (Equation 1), there are many systems for the RGB values converted from the tristimulus values, and it is necessary to match with the color system handled by the input data. Here, it is assumed that the input system such as a scanner or the sRGB system which is one of the RGB systems handled by the display is used as the input data.

The processing from the initial candidate setting means 6 to the rearrangement operation means 9 operates in accordance with the principle of an optimal solution search method called a genetic algorithm. For information on genetic algorithms, see, for example, “Genetic Algorithms in Searc.
h, Optimization and MachineLearning ”(David E. Go
ldberg, Addison Wesley) and others.
Excellent in global search ability, even if differential information of the evaluation function targeted for the optimal solution search problem cannot be obtained, if the evaluation value itself is obtained, the search can be performed, and only subjective evaluation of human judgment criteria It can be applied to the target problem without it, and its effectiveness is a method that has been reported so far. Specifically, the processing proceeds as follows.

<< Step 1 >> First, a set GM composed of n elements is considered based on the mm × mm color correction matrix M [k]. (I, j) component M [k] # i, j in M [k] (k = 1, ..., n), tristimulus value D (Dr, DGg, Db) and CMY signal (C, M , Y) is (Equation 4)
It is as follows. Here, the matrix dimension mm uses a non-linear second-order color correction matrix such as mm = 10, but in the case of a linear matrix such as (Equation 1), mm = 3,
In the case of a third-order color correction matrix, mm = 20. Each element VM [k] #i (i = 1,..., Mm) of VM [k] (referred to as a color correction vector here) in which this color correction matrix candidate M [k] is expressed in a mm × mm dimension vector × mm) is called a gene in relation to an organism, and the color correction vector VM [k] is called a chromosome. When using the genetic algorithm, first, the initial candidate setting means 6 performs a process of appropriately forming the initial set GM of the color correction vectors.

In the method 6, there are various methods for setting the chromosome, and the component M of the color correction matrix M [k] can be considered.
[k] #i is assumed to take a real value in the [c # i, d # i] section, and the value is converted to a bit string code of length len associated as shown in FIG. Color correction vector for the element
VM [k] is expressed in a form in which the coefficients M [k] #i converted into bit string codes are arranged in order. Here, the color correction vector is represented by the fixed-length bit string code as described above, but the method of expressing the color correction vector is not limited to this method. Conceivable.

According to an expression method using a bit string code,
Based on the uniform random number r in [0,1) and (Equation 5), the nn-th bit of the solution vector expressed in the form of a fixed-length bit string code
By calculating (nn = 1, .., len × mm × mm), an initial color correction vector set GM = [VM [k]] (k = 1,..., n) is set. Here, in (Equation 5), V [nn] represents the value of the nn-th bit from the lowest.

[0034]

(Equation 5)

The search for the optimum solution is started from the initial color correction vector set GM.

<< Step 2 >> The goodness of the solution of each element (chromosome) of the set VM is evaluated according to a preset evaluation scale, and the evaluation result of the color correction vector VM [k] is expressed as an evaluation value E [k]. Here, the preset evaluation scale is called an evaluation function (fitness function). This processing is performed by the evaluation value deriving means 7.
Do. Many evaluation functions E [k] can be considered. As shown in (Equation 6), nnn input color patch data sets Pi = (C, M, Y) i and read color patch data sets Ti = The color separation signal Di (Dr, Dg, Db) i obtained from (X, Y, Z) i is converted to each VM [k], that is, M [k].
Color chart data set Poi = (Co, Mo, Yo) i obtained using
The evaluation value E [k] for each color correction vector VM [k] is obtained by using the sum of squares of the error of Eq.

[0037]

(Equation 6)

In addition, the input color chart data set Pi = (C,
E [k] is the largest value among the square errors of the corrected color chart data set Poi = (Co, Mo, Yo) i obtained using M, Y) i and each VM [k], that is, M [k]. It is also conceivable.

<< Step 3 >> The fitness calculating means 8 calculates the fitness from the evaluation value obtained by the evaluation value deriving means 7 in order to check the fitness of each color correction vector. Conformity
There are various functions that can be used to derive f [k].
Here, f [k] = VE [k], and V is a positive real constant larger than E [k]. By doing so, the conversion can be made such that the smaller the evaluation value is, the larger the fitness is. A color correction vector with a low degree of matching is deleted from the current set of color correction vectors, and a color correction vector with a high degree of matching is selectively survived.

<Step 4> The color correction vector set selected in step 3 is subjected to a genetic rearrangement operation such as crossover or mutation to create a new color correction vector set. Note that the number of color correction vector individuals included in the color correction vector set is fixed here, but may be increased or decreased.

This processing is performed by the reassignment operation means 9. FIG.
The selective selection is performed by the roulette selection method of selecting a color correction vector with a probability proportional to the degree of fitness as shown in FIG.

(Roulette Selection Method) (i) Each color correction vector VM [k] (k = 1,..., N) belonging to the set GM
, And the sum F of the fitness of all color correction vectors.

(Ii) The selection probability h [k] that VM [k] is selected as a parent that creates a next-generation color correction vector is obtained as shown in (a). In order to assign this probability to the color correction vector VM [k], for example, the following method can be considered.

(Iii) The selection range I [k] of each color correction vector is
The section in [0,1) is assigned as shown in (b). Where [0,
The set RR of uniform random numbers ra [k] in 1) = (ra [1], ra [2], ..., ra [n])
Generate. nu satisfying ra [j] ∈I [k] (j, k = 1, ..., n)
By calculating a set of m [j] = I [k] Num = (num [1], num [2], ..., num [n]), the n color correction vectors corresponding to this Num are calculated. A pair will be selected.

By such a roulette selection method, the color correction vector VM [k] in the current color correction vector group GM is selected.

Next, the new color correction vector group GM thus obtained is subjected to crossover processing and mutation processing as shown in FIG. Crossover is an operation of creating a new color correction vector by replacing a part of a color correction vector represented by a finite symbol with a part of another color correction vector as shown in FIG. Mutation is an operation of changing a part of the components of the color correction vector selected from the color correction vector set to another symbol with a low probability as shown in FIG. The crossover process corresponds to a search at a position far away from the current solution vector, and the mutation corresponds to a search near the current color correction vector. Through these two processes, the genetic algorithm creates a new set of color correction vectors. Various methods have been proposed for processing such as crossover and mutation. In this embodiment, a one-point crossover process or a two-point crossover process of exchanging two points as shown in FIG. 3 is used. Further, as a mutation, a new color correction vector group GM obtained through the crossover process is subjected to a mutation process in which each bit constituting the color correction vector performs bit inversion with a low probability. At that time, the probability of performing the mutation was varied in half and the other half of the color correction vector group, so as to maintain the diversity of the color correction vectors.

As a convergence condition of the color correction matrix (or color correction vector) estimation processing by these genetic algorithms, it is determined whether the number of repetitions does not exceed an allowable number of repetitions. The process returns to the deriving means 7. The optimum color correction matrix is estimated by repeatedly executing the above-described processing steps until the convergence condition is satisfied. In this way, by preparing a large number of candidates globally and repeatedly performing symbolic rearrangement operations such as crossover and mutation that are difficult to trap in the local optimal solution (local minimum), problems with the conventional least squares method can be solved. Even if there is a contradiction in the learning color chart, the color correction matrix can be accurately estimated.
Also, by using the sum of the square errors of each component of the nnn input color chart data sets Pi = (C, M, Y) i prepared as an evaluation function, the color correction matrix can be obtained while balancing the three components. Can be optimized. As described in the conventional example, in the case of the least square method, a square error is obtained independently for three components of C, M, and Y, and a deviation of the color correction matrix component contributing to each of the components is set to 0 so that the color correction matrix is reduced. Since the coefficients are obtained independently, large color shifts are likely to occur when correcting colors that are slightly shifted from the sample. However, by using this method, it is possible to optimize the color correction matrix while balancing the three components. The color correction can be performed while maintaining the color balance with the surroundings even if there is an unlearned color input. Then, the color correction matrix M # best optimized by the genetic algorithm (or the color correction vector VM # best obtained by coding the bit string into a vector and stored in a vector) is held in 10 color correction coefficient databases. By performing arithmetic processing on the color separation signal D obtained by the data conversion means, the corrected color data 12 is output.

In this case, the color correction coefficient is held in a database of 10, and the coefficient and the color separation signal of 5 are stored in 11.
In order to reduce the calculation load of the higher-order color correction matrix M # ij, the higher-order matrix calculation is converted to a multidimensional table instead of the color correction coefficient database 10 in order to reduce the calculation load of the higher-order color correction matrix M # ij. Using a dimensional lookup table,
In the case of a color input that does not belong to the table, it is conceivable to use an interpolation operation in a triangular pyramid surrounded by four adjacent color chips in a color space as in the conventional method.

These processes are performed by software using a central processing unit (CPU) and a digital signal processor (DSP) used in a computer or the like according to the color image processing method according to the fourth embodiment of the present invention. The same can be realized in the processing.

(Second Embodiment) Next, a second embodiment of the present invention will be described.
A color image processing apparatus according to the embodiment will be described. In FIG. 6, reference numeral 60 denotes a gradient direction detecting means for detecting the gradient direction of the evaluation function E for a set GM of color correction vectors prepared from the current color correction matrix M # k. Is adjusted once with respect to the direction of the gradient direction vector delta # VM [k] obtained in the above, and new # VM [k] is converted into a bit string as described in the first embodiment of the present invention. This is a gradient direction adjusting means for converting into a code. The other parts are the same as in the first embodiment of the present invention, and will not be described.

The operation of the color image processing apparatus according to the second embodiment configured as described above will be described. As in the color image processing apparatus according to the first embodiment of the present invention, if there is no instruction to estimate a color correction matrix for input color data, the current color correction coefficient database stored in the 10 color correction coefficient database is used. Color correction matrix M # best, the color correction means 11 corrects the input color data into a signal subjected to color separation by the data conversion means 5, and is finally output as color corrected color data 12. .

On the other hand, when the instruction signal 2 is present, first, optimization estimation of the color correction matrix is performed as in the first embodiment of the present invention, and the color correction of the input color data is performed based on the value. It is executed. First, in step 5, the color system of the input color data is converted into the color separation signal D.
Then, based on this value, the color correction matrix is optimized by the genetic algorithm from 6 to 9.
The processes in 6, 7, 8, and 9 are performed in the same manner as in the first embodiment of the present invention. On the other hand, the gradient direction detecting means 6
0 and the gradient direction adjusting means 61, at this time, the color correction vector VM [k] obtained by the recombination operation of the genetic algorithm.
Is used to locally optimize the color correction vector using the gradient information of the evaluation value E [k] (gradient direction vector delta # VM [k]). This is a process for solving the drawback of the genetic algorithm, in which local information such as gradient information is not used, so that the estimation to the vicinity of the optimal solution is fast, but thereafter becomes very slow. In 60, each color correction vector VM
The gradient direction vector delta # VM [k] is obtained based on the evaluation function E [k] when [k] is used. Next, at 61, a line search is performed along the gradient direction vector in the direction in which the evaluation function E [k] decreases, and a color correction matrix M # lmin [that obtains the minimum value E # lmin [k] in that direction. k] is obtained using the color correction matrix M [k] when each VM [k] is returned to a real value as a starting point. Further, since each coefficient of the color correction matrix M # lmin [k] obtained here is a real value, each coefficient is converted into a binary bit string so that the rearrangement operation with the bit string in 9 can be performed. In this method, a color correction vector VM [k] is obtained by using the same method as in FIG.

As a gradient search method for obtaining the gradient based on the gradient information, a value obtained by multiplying a negative coefficient of the gradient direction vector by a steepest descent method instead of using the straight line search method in the gradient direction as described above. Is added to M [k], local optimization can be achieved, and optimization by the conjugate gradient method based on this gradient direction vector is also conceivable.

Naturally, the evaluation value E obtained by this VM [k]
Since [k] is different from E # lmin [k], the evaluation value deriving means 7
Derives the evaluation value E [k] for VM [k] again. And
Based on this value, the fitness calculating means 8 calculates the fitness f of each VM [k].
[k] is obtained, and based on this value, a new color correction vector VM [k] is estimated by performing selective selection, crossover, and mutation processing.

As described above, according to the embodiment,
By fusing the global optimization ability by the recombination operation of the genetic algorithm with the local optimization ability using gradient information such as the steepest descent method, the lack of the local search ability that was a problem with the genetic algorithm was solved. It is possible to make up for it, and the color correction accuracy for the learning color chart can be further improved.

Further, here, the color correction coefficient is held in a database of 10, and at 11 the coefficient and the color separation signal at 5 are stored.
In order to reduce the calculation load of the higher-order color correction matrix M # ij, the higher-order matrix calculation is converted to a multidimensional table instead of the color correction coefficient database 10 in order to reduce the calculation load of the higher-order color correction matrix M # ij. Using a dimensional lookup table,
In the case of a color input that does not belong to the table, it is conceivable to use an interpolation operation in a triangular pyramid surrounded by four adjacent color chips in a color space as in the conventional method.

These processes are performed by software using a central processing unit (CPU) and a digital signal processor (DSP) used in a computer or the like according to the color image processing method according to the fourth embodiment of the present invention. The same can be realized in the processing.

(Third Embodiment) Next, a color image processing apparatus according to a third embodiment of the present invention will be described. FIG. 8 shows a configuration of a color image processing apparatus according to the third embodiment. Numeral 80 denotes a neural network estimating means for learning the correspondence held in the database in 4 with the color correction matrix M # best estimated by the genetic algorithm as an initial state.

The operation of the color image processing apparatus according to the third embodiment configured as described above will be described with reference to the flowchart of FIG. Input color data 1
Is 10 if there is no color correction matrix estimation instruction signal 2.
Color corrected using the color correction matrix held in the
It is output as 12.

On the other hand, when the instruction signal 2 is present, the color correction matrix estimation is performed using the data in the learning color chart data set, but the processes from 5 to 9 are the same as those in the first embodiment of the present invention. Estimation processing is performed using a similar genetic algorithm. Then, the color correction matrix M # best obtained at the end of this processing has a three-layer hierarchical structure as shown in FIG. 10 and the output conversion function out of each unit is a sigmoid function as shown in FIG. Learning is performed based on the steepest descent method assuming the initial state of the coupling coefficient of the neural network. This is a process for compensating for the lack of local optimization capability of the genetic algorithm, as in the second embodiment of the present invention. When the number of repetitions, which is the convergence condition, is reduced and the evaluation function decreases to some extent globally, it is also possible to reduce the time for the estimation process by switching to estimation using a neural network.

Although the dimension mm of the color correction matrix has been described so far with mm = 10 (30 dimensions in terms of a vector), each coefficient M # ij of the color correction matrix M is represented by a three-layer pattern as shown in FIG. Coefficient W4,1, W4,2, W in the hierarchical structure of
4,3,. . . . . , Wm, 1, Wm, 2, Wm, 3,. . . . . , Wm +
In order to be assigned in correspondence with 3, m, it is necessary to prepare 3m × 2 color correction matrix coefficients. Where m is Figure 1
0 indicates the number of hidden layers in the hierarchical neural network. For this reason, the dimension of the color correction matrix tends to be large. However, when a neural network in which the output conversion function out of each unit is a sigmoid function as shown in FIG. It has been proved in principle that any nonlinear function approximation is possible. Therefore, more accurate color correction matrix approximation becomes possible.

In this manner, as in the color image processing apparatus according to the second embodiment of the present invention, the disadvantage of the genetic algorithm is compensated for, the color correction accuracy is improved, and the estimation speed is shortened.

In this case, the color correction coefficient is held in a database of 10 and the coefficient and the color separation signal of 5 are stored in 11.
In order to reduce the calculation load of the higher-order color correction matrix M # ij, the higher-order matrix calculation is converted to a multidimensional table instead of the color correction coefficient database 10 in order to reduce the calculation load of the higher-order color correction matrix M # ij. Using a dimensional lookup table,
In the case of a color input that does not belong to the table, it is conceivable to use an interpolation operation in a triangular pyramid surrounded by four adjacent color chips in a color space as in the conventional method.

These processes are performed by software using a central processing unit (CPU) and a digital signal processor (DSP) used in a computer or the like according to the color image processing method according to the fourth embodiment of the present invention. The same can be realized in the processing.

(Fourth Embodiment) Next, a color image processing apparatus according to a fourth embodiment of the present invention will be described. FIG. 11 shows a configuration of a color image processing apparatus according to the fourth embodiment, and FIG. 12 shows a flowchart of overall processing in the color image processing apparatus according to the fourth embodiment. FIG. 13 shows a configuration of a representative learning color chart selection means in the color image processing apparatus according to the fourth embodiment. FIG. 14 shows a representative learning color chart selection means in the color image processing apparatus according to the fourth embodiment. It shows the processing flow of the processing.
In FIG. 11, 110 converts the input color chart data set Pi = (C, M, Y) i generated in 3 into a uniform color space Qi = (L, a *, b *) i once, Is a color space dividing means for dividing into a plurality of areas, and 111 selects color chart data representative of the area based on the statistical distribution of the input color chart data sets belonging to each divided area. This is a representative learning color chart selection unit.

The representative learning color chart selection means 111 is configured as shown in FIG. Here, the representative learning color chart data set is derived by dividing the input color chart data distribution in the target area divided in the uniform color space such as the La * b * color space into a plurality of clusters. It is. There are many clustering methods, and here, the clustering method is configured based on a vector quantization (VQ) method. The vector quantization method is a method of dividing a vector space having a plurality of elements into a plurality of clusters based on a distribution state of a sample group.
6 illustrates a state in which a group of input vectors composed of two-axis components of β is clustered. The quantized representative vector is a vector representing a representative of each cluster when the target vector space is clustered into a plurality. Although there are many methods for this VQ, the method shown in FIG. 15 sets a cluster area frame as a vertical bisector of a line connecting adjacent quantized representative vectors.
It is also called the mean method. The representative learning color chart data obtained by the representative learning color chart selecting means 110 is exactly the data shown in FIG.
Corresponds to the quantized representative vector. Note that L represents lightness, and a * and b * correspond to hues.

First, the initialization means 130 sets the color data vector cv [k] = (vC [k], vM [k], vY [k]) (k =
1, ..., K) all belong to one cluster, and one representative color vector cc [1] = (cC [1], cM [1], cY
In [1]), set the centroid vectors of all cv [k]. Here, K represents the number of colors in the target color area, vC [k] represents the k-th C component, vM [k] represents the k-th M component, and vY [k] represents the k-th Y component. Where cC [1] is the C component of the representative color vector cc [1], and cM [1] is
M component and cY [1] represent Y component. Next, the division axis determination unit 131 determines a representative color vector cc [j] representing the cluster j in the current n clusters and n # m color vectors u [m] = (uC [m], uM [m], uY [m]
) (m = 1, ..., n # m)
(| uC [m]-cC [j] |, | uM [m]-cM [j] |, | uY [m]
-cY [j] |) and sum (tC = Σ | uC [m]-cC
[j] |, tM = Σ | uM [m]-cM [j] |, tY = Σ | uY [m]-c
Calculate Y [j] |) for each element. And tC, tM, tY
It is assumed that the current cluster can be divided in the direction of the axis having the largest value. The cluster dividing means 132
With the current representative color vector cc [j] as the center, two temporary representative color vectors dcc [j] and dcc are obtained in the axial direction obtained at 131.
Set [j + 1]. For example, when the L-axis is selected at 131, dcc [1] = (cC
[j] -gamma × tC, cM [j], cY [j]) and dcc [j + 1] = (cC [j] + gam
ma × tC, cM [j], cY [j]) are set at 132. Two temporary representative color vectors dcc from one representative color vector cc [j]
Since [j] and dcc [j + 1] are generated, and the processes at 131 and 132 are performed on all n clusters at the present time, the number of generated clusters is 2n. Note that this method is not unique, and other methods are possible. Then, 1
In the cluster representative determining means 33, 2n temporary representative color vectors dcc [i] newly obtained in 132 and the input color vector cv
Calculate the Euclidean distance disk [k, i] (i = 1, ..., 2n) between [k] and cv [ k] is assigned. Then, after each input color vector cv [k] can be assigned to a cluster, the center of gravity of the input color vector in the cluster i is set as the representative color vector cc [i] of the cluster i. The convergence determining means 134
The representative color vector cc [i] of the cluster converges based on the Euclidean distance between the representative color vector cc [i] of the cluster i obtained in and the provisional representative color vector dcc [i] obtained in 132. Is determined. If the convergence has not occurred, the processing is shifted to the cluster dividing means 132, and two representative temporary color vectors dcc [j] are obtained in the axial direction obtained at 131 around the current representative color vector cc [j]. It resets dcc [j + 1]. Specifically, dcc [j] = (cC [j-2] × gamma × tC, cM
[j], cY [j]) and dcc [j + 1] = (cC [j + 2] × gamma × tC, cM
[j], cY [j]) are set at 132 and obtained again at 133
Input color vector using 2n temporary representative color vectors dcc [i]
Return to the process of distributing cv [k] into 2n clusters. 13
If it is determined in step 4 that the convergence has been achieved, the number of clusters is reduced to a predetermined number of
Is determined, and if satisfied, the cluster finally obtained by the representative information output means 136 is output as representative learning color chart data. On the other hand, if the predetermined cluster division number mm is not satisfied, the process returns to the division axis determination means 131. In this configuration, the number of clusters finally obtained is always a multiple of two. However, at the stage of creating two clusters from one cluster at 131 and 132, the desired number of clusters is reduced to a desired number. At this point, it is also possible to exit the processing of 131 and 132 and move to the cluster representative determination means of 133.

As described above, the representative learning color chart selection means 11
1 and the processing therefor use VQ to divide the input color chart data in the divided area in the target color space into a plurality of clusters based on the distribution. However, this method is not unique, and instead of VQ, a method of simply dividing the maximum value or the minimum value of each color in order so that the histogram in each cluster becomes the same is possible. VQ method is used which has the feature of being able to perform cluster division with high accuracy according to the statistical distribution of. In addition, a method represented by a self-organizing neural network (for example, a Kohonen self-organizing network) can be used for other clustering.

Further, here, the center of gravity of the color vector cv [k] belonging to each cluster is set in 131 and 133 as the representative color vector representing each cluster.
It is also possible to divide the cluster by selecting the most appropriate color vector belonging to each cluster. Also, 13
When the color vector cv [k] is assigned to each cluster in step 3, the representative color vector cc [i] and the color vector cv [k] at the current time are
Euclidean distance dist [k, i] between is the smallest cluster i
Although it is assumed that cv [k] belongs, it is also possible to use the sum of absolute differences of the elements of the representative color vector cc [i] and the color vector cv [k] other than the Euclidean distance. And
The representative learning color chart data obtained in each area obtained here is output by an output device to be corrected, and its value is measured to prepare a reading color chart data set of the representative learning color chart. This is the learning color chart data set held at 4.

Based on these four data, a color correction matrix is estimated by a genetic algorithm through the same processing as in the first embodiment of the present invention. Then, using this result, the color correcting means 11 performs a color correcting process for the target output device of the input color data 1.

As described above, according to this embodiment,
Highly accurate color correction with a small number of samples by selecting multiple input color patch data sets that represent the entire input color range and corresponding read color patch data based on the statistical distribution of the entire input color range By converting the result into a 3D-LUT, a high-speed and high-precision color correction table can be created, and color calibration of an output device such as a printer can be easily performed.

In this case, the color correction coefficient is held in a database of 10, and the coefficient and the color separation signal of 5 are stored in 11.
In order to reduce the calculation load of the higher-order color correction matrix M # ij, the higher-order matrix calculation is converted to a multidimensional table instead of the color correction coefficient database 10 in order to reduce the calculation load of the higher-order color correction matrix M # ij. Using a dimensional lookup table,
In the case of a color input that does not belong to the table, it is conceivable to use an interpolation operation in a triangular pyramid surrounded by four adjacent color chips in a color space as in the conventional method.

Further, these processes are performed by software using a central processing unit (CPU) and a digital signal processor (DSP) used in a computer or the like according to the color image processing method according to the fourth embodiment of the present invention. The same can be realized in the processing.

(Fifth Embodiment) Next, a color image processing apparatus according to a fifth embodiment of the present invention will be described. FIG. 16 shows a configuration of a color image processing apparatus according to the fifth embodiment, and FIG. 17 shows a flowchart of overall processing in the color image processing apparatus according to the fifth embodiment. In FIG. 16, reference numeral 160 denotes a reference learning color chart generating means for generating a characteristic color chart in the input color space as input color chart data, and 161 denotes an interpolation learning color chart for filling interpolation color patches between the reference learning color charts. It is a generating means.

The operation of the color image processing apparatus according to the fifth embodiment configured as described above will be described. As in the color image processing apparatus according to the first embodiment of the present invention, if there is no instruction to estimate a color correction matrix for input color data, the current color correction coefficient database stored in the 10 color correction coefficient database is used. Color correction matrix M # best, the color correction means 11 corrects the input color data into a signal subjected to color separation by the data conversion means 5, and is finally output as color corrected color data 12. .

On the other hand, when the instruction signal 2 is present, the optimization of the color correction matrix is first performed in the same manner as in the first embodiment of the present invention, and the color correction of the input color data is performed based on the value. It is executed. First, in step 5, the color system of the input color data is converted into the color separation signal D.
Then, based on this value, the color correction matrix is optimized by the genetic algorithm from 6 to 9. The method of creating the learning color chart data set 4 used for the estimation is different from that of the first embodiment of the present invention. First, in the output color space, red, green, blue, cyan,
A pure color such as magenta or yellow and a characteristic color (skin color, etc.) in human vision are output as reference input color chart data, and the output result is measured by a spectrophotometer to be used as reference read color chart data. . 160 prepares these two combinations as a reference learning color chart data set. The interpolation learning color chart generation means 161 generates an interpolation learning color chart between the reference learning color chart data. Various methods can be considered as a method of creating the interpolation learning color chart data. Here, as shown in FIG. 18, a spline function often used in free-curve drawing and approximation of a commercially available graphic tool such as Photoshop is used. Use an approximation technique. By performing this process on the reference input color patch data and also on the reference read color patch data, it is possible to prepare the read color patch data in addition to the input color patch data associated therewith. . In this case, as a function for generating the interpolation learning color chart, it can be expressed by superposition of trigonometric functions, or can be expressed by a multidimensional polynomial. It is also possible to obtain an interpolation learning color chart generation function by learning on some learning color charts in advance by the neural network used in the second embodiment of the present invention. Can reduce the influence on the degree of generation of the interpolation learning color chart due to the selection by.

As described above, according to the fifth color image processing apparatus of the present invention, the read color patch data corresponding to a plurality of characteristic input color patch data sets is selected, and the selected input color patch data is selected. A high-precision color correction can be realized with a small number of samples by preparing an input color patch data set sufficient for color correction estimation and a corresponding read color patch data set by successfully interpolating between data and between read color patch data. Can be. In order to reduce the calculation load of the higher-order color correction matrix M # ij, a three-dimensional look-up table in which the higher-order matrix calculation is converted into a multi-dimensional table is used instead of the color correction coefficient database 10. In the case of a color input that does not belong, it is conceivable to use an interpolation operation in a triangular pyramid surrounded by four adjacent color chips in a color space as in the conventional method.

Further, these processes are performed by software using a central processing unit (CPU) and a digital signal processor (DSP) used in a computer or the like according to the color image processing method according to the fourth embodiment of the present invention. The same can be realized in the processing.

(Sixth Embodiment) Finally, the sixth embodiment of the present invention
An image enlarging apparatus according to the embodiment will be described. FIG. 19 illustrates a configuration of a color image processing apparatus according to the sixth embodiment, and FIG. 20 illustrates a flowchart of overall processing in the color image processing apparatus according to the sixth embodiment.

As is apparent from FIG. 19, the present invention provides a learning color chart generating means 3 similar to the fourth embodiment of the present invention.
Subdivides the input color patch data set prepared in the above into a uniform color space such as La * b * space, and creates representative input color patch data representing that area based on the statistical distribution of the input color patch data sets in each subdivided area And a representative reading color patch data set obtained by measuring the color of the representative input color patch data set and the result of printing the data on an output device as used in the fifth embodiment of the present invention. It is characterized by a process of creating interpolation learning color chart data for interpolating between representative learning color chart data sets as a set.

In this way, when a method of easily performing color correction calibration of an output device with a small amount of learning color chart data as in the fourth embodiment of the present invention is used,
By compensating for the delicate learning color chips that are lost during the generation of the representative learning color chips with the interpolation learning color chips, it is possible to improve the accuracy of the color correction matrix generated by pushing the representative learning color chips into the representative learning color chip data.

As described above, according to the sixth color image processing apparatus of the present invention, based on the statistical distribution of the entire gamut of the input color gamut, the read color for a plurality of input color chart data sets representing the entire gamut of the input color gamut. A small number of output samples can be obtained by selecting the form data and preparing a sufficient input color form data set and a corresponding read color form data set for estimating the color correction matrix between the selected input color form data and the read color form data. Color correction with sufficient accuracy can be realized. Also, by interpolating between the prepared representative learning color chart data, it is possible to compensate for the color chart information that has been dropped when the data is consolidated into the representative learning color chart, and finally a high-precision color correction matrix Can be estimated. The present invention has a great effect in that color calibration of an output device such as a printer can be easily performed.

In order to reduce the calculation load of the higher-order color correction matrix M # ij, a three-dimensional lookup table in which the higher-order matrix calculation is converted into a multi-dimensional table is used instead of the color correction coefficient database 10. In the case of a color input that does not belong to the table, it is conceivable to use an interpolation operation in a triangular pyramid surrounded by four adjacent color chips in a color space as in the conventional method.

Further, these processes are performed by software using a central processing unit (CPU) and a digital signal processor (DSP) used in a computer or the like according to the color image processing method according to the fourth embodiment of the present invention. The same can be realized in the processing.

[0085]

As described above, the first color image processing method and the color image processing apparatus according to the present invention use the evaluation function derived between the input color patch data and the corresponding read color patch data in advance. By estimating the color correction matrix with the optimal evaluation value using a genetic algorithm that performs symbolic rearrangement processing for each candidate in the prepared multiple color correction matrix candidate group, highly accurate even for unlearned color chips The color correction can be realized, and there is an effect of reducing the color shift after the color correction generated for the input data slightly shifted from the learning color chart.

The second color image processing method and the color image processing apparatus of the present invention are used to estimate a color correction matrix for optimally converting an input color patch data set and a corresponding read color patch data set. By blending the genetic algorithm used in the first aspect of the present invention with the method using gradient information of the evaluation value, the lack of local search capability, which has been a problem in the genetic algorithm, can be compensated. This leads to an improvement in color correction accuracy for votes.

The third color image processing method and the color image processing apparatus according to the present invention, when estimating the optimal color correction matrix for the correspondence between the input color patch data set and the read color patch data set, correspond to the present invention. After estimating the genetic algorithm used in the first embodiment, the neural network technique is used again for estimation, thereby improving the color correction accuracy and shortening the estimation speed as in the second invention. Connect.

The fourth color image processing method and the color image processing apparatus of the present invention correspond to a plurality of input color chart data sets representing the entire input color range based on the statistical distribution of the entire input color range. By selecting reading color chart data, high-precision color correction can be realized with a small number of samples, and by converting this result into a 3D-LUT, a high-speed and high-precision color correction table can be created. Calibration of device colors can be easily performed.

The fifth color image processing method and the color image processing apparatus of the present invention select read color patch data corresponding to a plurality of input color patch data sets that are visually important in the input color range and select the read color patch data. Interpolation between input color patch data and read color patch data to prepare a sufficient input color patch data set and a corresponding read color patch data set for color correction estimation enables highly accurate color correction with a small number of samples. By converting this color correction result into a 3D-LUT, it is possible to easily and accurately calibrate the color of the output device simply by preparing a color chart sample which is not as in the fourth invention. effective.

A sixth color image processing method and a color image processing apparatus according to the present invention read a plurality of input color patch data sets representing the entire input color range based on the statistical distribution of the entire input color range. By selecting color patch data and preparing a sufficient input color patch data set and a corresponding read color patch data set between the selected input color patch data and estimating the read color patch data color correction matrix, the number of samples can be reduced. High-precision color correction can be realized, and by converting this color correction result to 3D-LUT, color calibration of output devices such as printers can be easily performed, and the color correction accuracy for unlearned color charts can be further improved. Has the expected effect of improvement.

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating a configuration of a color image processing apparatus according to a first embodiment of the present invention.

FIG. 2 is a flowchart illustrating a processing flow of a color image processing method according to the first embodiment of the present invention;

FIG. 3 is a conceptual diagram schematically illustrating conversion from a value in a real-valued section to a fixed-length bit string code in the color image processing apparatus according to the first embodiment of the present invention;

FIG. 4 is a schematic diagram illustrating a roulette selection method used for selective selection of the color image processing apparatus according to the first embodiment of the present invention.

FIG. 5 is a conceptual diagram of a rearrangement operation process in a rearrangement operation unit in the color image processing apparatus according to the first embodiment of the present invention;

FIG. 6 is a block diagram illustrating a configuration of a color image processing apparatus according to a second embodiment of the present invention.

FIG. 7 is a flowchart illustrating a processing flow of a color image processing method according to a second embodiment of the present invention;

FIG. 8 is a block diagram illustrating a configuration of a color image processing apparatus according to a third embodiment of the present invention.

FIG. 9 is a flowchart illustrating a processing flow of a color image processing method according to a third embodiment of the present invention.

FIG. 10 is a schematic diagram illustrating a neural network used for estimating a color correction matrix estimated by a color image processing apparatus according to a third embodiment of the present invention.

FIG. 11 is a block diagram illustrating a configuration of a color image processing apparatus according to a fourth embodiment of the present invention.

FIG. 12 is a flowchart illustrating an overall processing flow of a color image processing method according to a fourth embodiment of the present invention;

FIG. 13 is a block diagram illustrating a configuration of a representative learning color chart selection unit in a color image processing apparatus according to a fourth embodiment of the present invention.

FIG. 14 is a flowchart illustrating a process flow of a representative learning color chart selection process in the color image processing apparatus according to the fourth embodiment of the present invention.

FIG. 15 is a conceptual diagram of a vector quantization method used in a representative learning color chart selection unit in a color image processing apparatus according to a fourth embodiment of the present invention.

FIG. 16 is a block diagram illustrating a configuration of a color image processing apparatus according to a fifth embodiment of the present invention.

FIG. 17 is a flowchart illustrating a processing flow of a color image processing method according to a fifth embodiment of the present invention.

FIG. 18 is a conceptual diagram showing a concept of an interpolation learning color chart creation method by an interpolation learning color chart creation means in a color image processing apparatus according to a fifth embodiment of the present invention.

FIG. 19 is a block diagram illustrating a configuration of a color image processing apparatus according to a sixth embodiment of the present invention.

FIG. 20 is a flowchart illustrating a processing flow of a color image processing method according to a sixth embodiment of the present invention.

FIG. 21 is a block diagram illustrating a configuration of an image processing apparatus according to a conventional example.

FIG. 22 is a diagram showing an example of an input color patch data set and a read color patch data set.

[Explanation of symbols]

 REFERENCE SIGNS LIST 1 input color data 2 instruction signal 3 learning color chart generation means 4 learning color chart data set 5 data conversion means 6 initial candidate setting means 7 evaluation value derivation means 8 fitness calculation means 9 recombination operation means 10 color correction coefficient database 11 color correction Means 12 Modified color data 60 Gradient direction detecting means 61 Gradient direction adjusting means 80 Neural network estimating means 110 Color space dividing means 111 Representative learning color chart selecting means 130 Initializing means 131 Dividing axis determining means 132 Cluster dividing means 133 Cluster representative determining means 134 Convergence judgment means 135 Cluster division end judgment means 136 Representative color chart output means 160 Reference learning color chart generation means 161 Interpolation learning color chart generation means 210 Least square estimation means

 ──────────────────────────────────────────────────続 き Continuing on the front page (72) Inventor ▲ Kawa ▼ Yasuhiro Hara 1006 Kadoma Kadoma, Osaka Prefecture Inside Matsushita Electric Industrial Co., Ltd. (72) Toshiharu Kurosawa 1006 Odakadoma Kadoma, Osaka Prefecture Matsushita Electric Industrial In-company F-term (reference) 5B057 CA01 CA08 CA12 CA16 CB01 CB08 CB12 CB16 CC01 CE17 DB02 DB06 DB09 DC25 5C077 MM27 MP08 PP32 PP33 PP37 PP47 PQ08 PQ12 PQ15 PQ23 RR19 5C079 HB02 LA02 LA10 LA24 LA11 MA03 NA03 PA05

Claims (15)

[Claims] 1. A color image processing method for creating a color correction matrix from an input color patch data set adapted to the entire area of an input color range and a corresponding read color patch data set and correcting the color of color input data. Based on the evaluation function derived between the color chart data and the corresponding read color chart data, it is optimized by performing symbolic rearrangement processing on each candidate in the multiple color correction matrix candidate group prepared in advance. A color image processing method for estimating a color correction matrix having an evaluation value and converting the digital image data into color-corrected image data to be output by an output device using the color correction matrix. 2. A color image processing method for creating a color correction matrix from an input color patch data set adapted to the entire area of an input color range and a corresponding read color patch data set and correcting the color of color input data. Based on the evaluation function derived between the color chart data and the corresponding read color chart data and its gradient information, a symbolic rearrangement process is performed for each candidate in a plurality of color correction matrix candidate groups prepared in advance. A color correction matrix having an optimum evaluation value is estimated by using a gradient search method for searching for a gradient based on the gradient information, and the color to be output by an output device from digital image data using the color correction matrix. A color image processing method for converting to corrected image data. 3. A color image processing method for creating a color correction matrix from an input color patch data set adapted to the entire area of an input color range and a corresponding read color patch data set and correcting the color of color input data. Based on the evaluation function derived between the color chart data and the corresponding read color chart data, an appropriate evaluation value is obtained once by using symbolic rearrangement processing for each candidate in a plurality of color correction matrix candidate groups prepared in advance. Estimating the color correction matrix with, and applying the neural network method that realizes the correspondence between the input color patch data and the corresponding read color patch data by learning based on the gradient information of the evaluation value A color correction matrix having an evaluation value is estimated, and the color correction matrix is used to convert the digital image data into color-corrected image data to be output by an output device. Color image processing method. 4. A color image processing method for creating a color correction matrix from an input color patch data set adapted to the entire area of an input color range and a corresponding read color patch data set and correcting the color of color input data. Based on the statistical distribution of the entire color gamut, select a plurality of input color patch data sets representing the entire input color gamut and the corresponding read color patch data, and select the read color corresponding to the selected input color patch data Based on the evaluation function derived between the vote data, a color correction matrix having an optimal evaluation value is obtained by performing symbolic rearrangement processing on each candidate in a plurality of color correction matrix candidate groups prepared in advance. A color image processing method for estimating and converting digital image data into color-corrected image data to be output on an output device using the color correction matrix. 5. A color image processing method for preparing a color correction matrix from an input color patch data set adapted to the entire area of an input color range and a corresponding read color patch data set and correcting the color of color input data. It is necessary to select the read color patch data corresponding to a plurality of input color patch data sets characteristic in the color range, and to interpolate between the selected input color patch data and between the read color patch data to estimate the color correction matrix. Prepare an input color patch data set and a corresponding read color patch data set, and prepare a plurality of colors prepared in advance based on the evaluation function derived between the prepared input color patch data and the corresponding read color patch data. By performing symbolic rearrangement processing on each candidate in the candidate group of the correction matrix, a color correction matrix having an optimal evaluation value is estimated, and the color correction matrix is calculated. Color image processing method for converting the image data after the color correction to be output by the output device from the digital image data are. 6. A color image processing method for creating a color correction matrix from an input color patch data set adapted to the entire area of an input color range and a corresponding read color patch data set and correcting the color of color input data. Based on the statistical distribution of the entire color gamut, select read color patch data for a plurality of input color patch data sets representing the entire input color gamut, and select between selected input color patch data and read color patch data By preparing an input color patch data set and a corresponding read color patch data set necessary for color correction matrix estimation by interpolating the space, an evaluation derived between the prepared input color patch data and the corresponding read color patch data Based on the function, a symbol correction process is performed on each candidate in the candidate group of a plurality of color correction matrices prepared in advance to obtain a color correction matrix having an optimal evaluation value. A color image processing method for estimating the risk and converting the digital image data into color-corrected image data to be output by an output device using the color correction matrix. 7. An evaluation function derived between input color patch data corresponding to the entire area of the input color range and read color patch data corresponding thereto, the prepared input color patch data and read color patch data corresponding to the prepared input color patch data. The color image processing method according to any one of claims 1 to 6, wherein a sum of square errors between estimated color chart data estimated by color correction matrix approximation is used. 8. A learning color patch data generating means for printing an input color patch data set based on an instruction signal and reading the color information, and a read color patch data set corresponding to the input color patch data set. A learning color chart database; data conversion means for converting the read color chart data into data for color correction; initial candidate setting means for setting an initial candidate group of a color correction matrix; Evaluation value deriving means for deriving an evaluation value of each candidate in the candidate group of the color correction matrix, a fitness calculating means for calculating the fitness of each candidate obtained by the evaluation value deriving means, and the fitness calculation Means for performing a rearrangement operation of the current candidates based on the degree of matching obtained by the means to generate a new candidate set; and a color image processing apparatus comprising: Apparatus. 9. A learning color patch data generating means for printing an input color patch data set based on an instruction signal and reading color information thereof, and a read color patch data set corresponding to the input color patch data set. A learning color chart database; data conversion means for converting the read color chart data into data for color correction; initial candidate setting means for setting an initial candidate group of a color correction matrix; Gradient direction detecting means for deriving a gradient component of the evaluation value of each candidate in the candidate group of the color correction matrix, gradient direction searching means for searching in the gradient direction of each candidate obtained by the evaluation value gradient deriving means, An evaluation value deriving unit that derives an evaluation value of each candidate in the candidate group of the current color correction matrix using the learning color chart database; and A fitness calculation means for calculating the fitness of the candidate, and a reclassification operation means for generating a new candidate set by performing a regrouping operation of the current candidates based on the fitness obtained by the fitness calculation means, And a color image processing apparatus. 10. A learning color chart generating means for printing an input color chart data set based on an instruction signal and reading the color information, and a read color chart data set corresponding to the input color chart data set. A learning color chart database; data conversion means for converting the read color chart data into data for color correction; initial candidate setting means for setting an initial candidate group of a color correction matrix; Evaluation value deriving means for deriving an evaluation value of each candidate in the candidate group of the color correction matrix, a fitness calculating means for calculating the fitness of each candidate obtained by the evaluation value deriving means, and the fitness calculation Means for performing a rearrangement operation of the current candidates based on the fitness obtained by the means to generate a new candidate set; and Neural network estimating means for applying a neural network method for realizing the correspondence between the input color patch data and the corresponding read color patch data by learning with the optimal color correction matrix as an initial state. Color image processing device. 11. A learning color chart generating means for printing an input color chart data set based on an instruction signal and reading color information thereof, and reading a read color chart data set corresponding to the input color chart data set into a plurality of areas. A color space dividing unit that divides the input color patch data set belonging to each area obtained by the color space dividing unit and a read color patch data set corresponding to
Representative learning color patch selection means for selecting a plurality of input color patch data sets representing the entire input color range and corresponding read color patch data based on the statistical distribution of the entire input color range, and the selected input A learning color chart database composed of a color chart data set and a corresponding read color chart data set; data conversion means for converting the read color chart data into data to be subjected to color correction; and setting an initial candidate group of a color correction matrix. Initial candidate setting means for performing evaluation, evaluation value deriving means for deriving an evaluation value of each candidate in the current color correction matrix candidate group using the learning color chart database, and each candidate obtained by the evaluation value deriving means. A relevance calculating means for calculating the relevance of the data, and a rearrangement operation for generating a new candidate set by performing a rearrangement operation on the current candidates based on the relevance obtained by the relevance calculating means. Color image processing apparatus comprising: the stage, to be more configuration. 12. A learning color chart generating means for printing an input color chart data set based on an instruction signal and reading the color information; and a reading color chart data set corresponding to the input color chart data set. Reference learning color chart generation means for selecting reading color chart data corresponding to the input color chart data set; and interpolation learning color chart for obtaining an interpolation value between the selected input color chart data set and the corresponding reading color chart data set. Generating means, a learning color patch database comprising input color patch data and read color patch data obtained by the reference learning color patch generating means and the interpolation learning color patch generating means, and correcting the read color patch data. Data conversion means for converting into target data; initial candidate setting means for setting an initial candidate group of a color correction matrix; and a current color correction matrix using the learning color chart database. Evaluation value deriving means for deriving an evaluation value of each candidate in the candidate group of; a fitness degree calculating means for calculating fitness of each candidate obtained by the evaluation value deriving means; And a rearrangement operation unit for generating a new candidate set by performing a rearrangement operation of the current candidates based on the degree of matching. 13. A learning color chart data generating means for printing an input color chart data set based on an instruction signal and reading color information thereof, and reading a read color chart data set corresponding to the input color chart data set into a plurality of areas. A color space dividing unit that divides the input color patch data set belonging to each area obtained by the color space dividing unit and a read color patch data set corresponding to
A representative learning color chart selecting means for selecting a plurality of input color chart data sets representing the entire input color range and corresponding read color chart data based on a statistical distribution of the entire input color range; Interpolation learning color chart generation means for obtaining an interpolation value between each input color patch data set selected as data and the corresponding read color patch data set; and the representative learning color patch selection means and interpolation learning color patch generation means. A learning color patch database including a read color patch data set corresponding to an input color patch data set in each area, data conversion means for converting the read color patch data into data of a color correction target, and a color correction matrix. Initial candidate setting means for setting an initial candidate group of, and evaluation value deriving means for deriving an evaluation value of each candidate in the candidate group of the current color correction matrix using the learning color chart database. A fitness calculating means for calculating the fitness of each candidate obtained by the evaluation value deriving means; and a new candidate by performing a recombining operation of the current candidate based on the fitness obtained by the fitness calculating means. A color image processing apparatus comprising: a rearrangement operation unit for generating a set; 14. An evaluation function deriving between input color patch data corresponding to the entire area of the input color range and read color patch data corresponding to the entire input color patch data in the evaluation value deriving means, which corresponds to all prepared input color patch data. 14. The color image processing apparatus according to claim 8, wherein a sum of square errors between the read color patch data and the estimated color patch data estimated by the color conversion matrix approximation is used. 15. The color image processing apparatus according to claim 8, wherein a genetic algorithm represented by crossover, mutation, and selection is used as the recombination operation means.