JPH11266371A - Color image processor - Google Patents

Abstract

PROBLEM TO BE SOLVED: To obtain a color image processor which realizes color reproducibility corresponding to a human visual sensation characteristic by weighting the error of each chromaticity signal in plural chromaticity signals in the calculation of the whole errors of plural chromaticity signals in a neural network. SOLUTION: A 1st color converting part 3 sets weighted factor values KL, Ka , Kb to the error of each chromaticity in three chromaticity signals L*a*b* in calculating energy E of three chromaticity signals from a color signal obtained from an input device 1 in a neutral network. At least one of the three chromaticity signals L*a*b* is a lightness component or a luminance component. Then, the part 3 can adjust weight to the error of each chromaticity signal L*a*b* about the lightness component and the luminance component which are sensitive to human visual sensation. As a result of this, it is possible to realize color reproducibility that corresponds to human visual sensation characteristic.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a device for performing digital image processing, for example, a digital color copier or a color printer, which performs color conversion processing on a color signal obtained from an input device and outputs the color signal to an image output device. And the like.

[0002]

2. Description of the Related Art Conventionally, a color image processing apparatus such as a color copying machine or a color printer which reads a color original image by an input device such as a color scanner and outputs a duplicate image of the color original image by an image output device such as a color printer. There is. In such a color image processing apparatus, even if the input image data is directly input to the image output apparatus, in most cases, the input image data is output as a duplicate image having a color different from the original image.

Various color correction techniques have been proposed in order to obtain a reproduced image whose color is faithfully reproduced from the original image. Representative color correction techniques include the color correction processing by digitization using polynomial regression analysis, which is described in “Color Correction Techniques for Faithful Color Reproduction”, Vol. 29, No. 3, 1990 (1990). A color correction process by non-formulaization using three-dimensional interpolation using a lookup table and a color correction process by non-formulaization using a neural network are described.

[0004] Among the above-described color correction processes, in particular, in a color correction process using a neural network, even if a large number of colors are used in an original image, it is possible to perform a desired color correction simply by giving color data. It is possible. In this method, color data to be sampled is given to the neural network, and the neural network itself learns spontaneously so that the target color correction data is obtained, and the neural network itself finds a processing procedure of the color correction. This is because

[0005] In addition, the neural network is
Since a large amount of data can be processed in parallel, it is suitable for color correction processing when the number of colors of the original image is large.

In a color image processing apparatus for performing a color correction process using the above-described neural network, as shown in FIG. 5, a document is read by an input device 51 such as a scanner.
After converting the color separation signals into RGB, that is, red, green, and blue, the color separation signals are input to the first color conversion unit 52. Here, the first color conversion unit 52 performs color conversion using the following neural network, that is, a neural network.

That is, a neural network is a circuit network imitating a neural cell network of the brain, and units corresponding to nerve cells, which are basic elements of the brain, are connected to each other to form a network. This neural network includes an input layer 11, a first intermediate layer 12, a second intermediate layer 13, and an output layer 14 as shown in FIG.
And three units, for example, five units, five units, and three units, respectively, in order to correspond to nerve cells. The output value is quantized by the quantization unit 17 to become a final 8-bit output signal (L ^* a ^* b ^* ).

The function of each unit 15... In the above-mentioned neural network in the conventional intermediate layer will be described with reference to FIG.

An input u _i (i = 1 to n) is an output value from each unit 15... Of the previous layer, and a weight value w _i (i = 1 to n).
n) is a constant obtained by a predetermined method described later.
These weight values w _i are stored in the first storage device 55 shown in FIG. Input u _i and the weight value w _i The above by multiplying unit 61 ... and the addition unit 62, the signal u _p shown in Equation 1.

[0010]

(Equation 1)

[0011] The signal u _p is converted by the non-linear calculation unit 63 into a signal u _out. Here, the operation performed by the non-linear operation unit 63 is as shown in Expression 2.

[0012]

(Equation 2)

The above is the function of each unit 15. A plurality of units 15 are combined to form a neural network.

Incidentally, for the optimization of the weight values w _{1 to} w _n , a back propagation method or the like known in neural network theory is generally used.

In the back propagation method, a desired output with respect to an input is given in advance as a teacher signal, and a function converted from a difference between an actual output and the teacher signal is defined as an energy E. In a predetermined output set, the energy E is Each weight value w _i is changed until it decreases or converges to 0 or saturation, that is, to a certain value.

Here, in order to apply the back propagation method to the neural network, conventionally, first, n color patches whose values of L ^* a ^* b ^* are known are input to the input device 51. Read, n L ^* a ^* b ^*
Get the value of the signal. Then, the value of each of the n L ^* .a ^* .b ^* signals is used as a teacher signal. At this time, the energy E is expressed by Expression 3.

[0017]

(Equation 3)

Here, L ^* .a ^* .b ^* is a teacher signal,
L ^* 'a ^* ' b ^* 'is a predicted value by a neural network. Therefore, optimization can be performed by learning in which the energy E is reduced or becomes zero or saturated. Then, the obtained weight values w _{1 to} w _n are stored in the first storage device 55 as shown in FIG.

The input color signals (RGB) converted by the first color conversion section 52 in this way become device-independent color signals (L ^* a ^* b ^* ), and thereafter, the second storage. Based on the data stored in the device 56, the data is converted into color signals such as CMY, that is, yellow, magenta, and cyan, by the second color conversion unit 53.

As the color conversion of the second color conversion unit 53, similarly to the first color conversion unit 52, the desired CM is input using the color sample CIEL ^* a ^* b ^* of the image output device 54 as an input.
A neural network that will be Y is constructed.

These similar techniques are disclosed in, for example,
No. 181520 and JP-A-8-204973.

These CMY signals are sent to the image output device 54 and output.

[0023]

However, in the above-mentioned conventional color image processing apparatus, since the energy E is calculated by the equation 3, an error is inevitably generated after the learning of the neural network is completed. In particular, this error is
This is likely to occur in a portion where the nonlinearity is high.

That is, the above-mentioned error is evaluated by the color difference ΔEab, and the error is equally allocated to L ^* a ^* b ^* .

However, generally, human visual characteristics are L
^* (Lightness) is sensitive and the error of L ^* is a ^*
And b ^* (saturation and hue) are preferably smaller.

Therefore, the error function represented by the above equation 3 cannot cope with this point, and in an actual print sample, inversion of lightness may occur, and the color reproducibility according to human visual characteristics is not sufficient. There is a problem that.

The present invention has been made in view of the above-mentioned conventional problems, and has as its object to provide a color image processing apparatus capable of realizing color reproducibility according to human visual characteristics.

[0028]

According to a first aspect of the present invention, there is provided a color image processing apparatus for performing a color conversion process on a color signal obtained from an input device and outputting an image as a color image. A color image processing device for outputting to the device, a color conversion device for converting a color signal obtained from the input device into three chromaticity signals by a neural network; When calculating the total error of the chromaticity signal, weighting is performed on the error of each of the chromaticity signals in the above three chromaticity signals.

According to the above invention, when the color conversion means calculates the total error of the three chromaticity signals by the neural network using the color signal obtained from the input device, the color conversion means includes the three chromaticity signals. Weighting is performed on the error for each chromaticity signal.

That is, in general, human visual characteristics are sensitive to lightness, and it is desirable that the error of the lightness component is smaller than the error of other chroma and hue components. However, in the conventional method of calculating the total error using the neural network, the error of each chromaticity signal is simply added, and as a result, the color reproducibility is not sufficient.

However, in the present invention, when calculating the total error of the three chromaticity signals, the error of each of the three chromaticity signals is weighted.

Therefore, it is possible to calculate a total error in consideration of the degree of influence of each chromaticity signal, and as a result, to provide a color image processing apparatus capable of realizing color reproducibility according to human visual characteristics. Can be.

According to a second aspect of the present invention, there is provided a color image processing apparatus according to the first aspect, wherein at least one of the chromaticity signals is a lightness component or a luminance component. It is characterized by having.

According to the above invention, the color conversion means calculates the total error of the three chromaticity signals of the color signals obtained from the input device by using a neural network. One is a brightness component or a luminance component.

Therefore, the color conversion means can adjust the weight for the error of each chromaticity signal with respect to the lightness component or the luminance component that is sensitive to human vision.

As a result, it is possible to provide a color image processing apparatus capable of realizing color reproducibility according to human visual characteristics.

According to a third aspect of the present invention, there is provided a color image processing apparatus as set forth in the second aspect, wherein the weighting coefficient value for the brightness component or the luminance component is: The other two components, namely the lightness component, are characterized by being larger than the hue and saturation, and the luminance component are greater than the coefficient values for weighting the hue and the color saturation.

According to the above-described invention, the coefficient values for weighting the lightness component or the luminance component are different from those of the other two components, namely, the hue and saturation for the lightness component, and the hue and color saturation for the luminance component. Is larger than the coefficient value of the weighting for, the lightness component or the luminance component that is sensitive to human vision can be corrected to something closer to reality.

That is, in various chromaticity space coordinates, the lightness component or the luminance component usually has a larger spatial distance than the other components. Therefore, in order to bring the brightness component or the luminance component closer to the real one, as described above,
The weighting coefficient value for the lightness component or the luminance component,
It is necessary to make it larger than the weighting coefficient value for the other two components.

As a result, it is possible to provide a color image processing apparatus capable of realizing color reproducibility according to human visual characteristics.

According to a fourth aspect of the present invention, there is provided a color image processing apparatus according to the second aspect, wherein the chromaticity signal is CIEL ^* a ^*.
b ^* .

According to the above invention, the chromaticity signal is CIEL
^* a ^* b ^* , that is, L ^* a ^* defined by the International Commission on Illumination
b ^* and L ^* , which is a lightness component that is sensitive to human vision, the color conversion means adjusts the weight for the error in the chromaticity signal for this lightness component Can be.

As a result, it is possible to provide a color image processing apparatus capable of realizing color reproducibility according to human visual characteristics.

According to a fifth aspect of the present invention, there is provided a color image processing apparatus as set forth in the fourth aspect, wherein L of CIEL ^* a ^* b ^* is used.
The weighting coefficient value for ^* is the other two components a ^*.
It is characterized in that it is larger than the coefficient value for weighting b ^* .

According to the above invention, the chromaticity signal is CIEL
^* a ^* b ^* , which includes L ^* which is a lightness component that is sensitive to human vision, and the color conversion means
The IEL ^* a ^* b ^* is adjusted so that the coefficient value for weighting L ^* is larger than the coefficient value for weighting the other two components a ^* and b ^* .

Therefore, the color signal obtained from the input device is represented by C
When color conversion processing is performed on a chromaticity signal of IEL ^* a ^* b ^* and output to an image output device as a color image, among the unavoidable calculation errors conventionally existing, it is sensitive to human vision. The lightness component can be corrected to something closer to reality.

As a result, it is possible to provide a color image processing apparatus capable of realizing color reproducibility according to human visual characteristics.

[0048]

DESCRIPTION OF THE PREFERRED EMBODIMENTS One embodiment of the present invention will be described below with reference to FIGS.

In this embodiment, for example, in a color image processing apparatus such as a color copying machine or a color printer, a color original image is represented as a representative color of red (R), green (G), and blue (B) by an input device. An input signal (hereinafter, referred to as an RGB signal) is input, and the input data of the original image is input to the Commission Internationale d'Icla (CIE).
A description will be given of a case in which the color of an input image is corrected by conversion into an L ^* a ^* b ^* color system defined in irage)).

In a color copying machine, for example, as a color image processing apparatus according to the present embodiment, as shown in FIG. 2, an input device 1 such as a scanner for inputting an original, and a γ correction unit 2
And a first color conversion unit 3, a second color conversion unit 4, and an image output device 5 that outputs the original image subjected to the color conversion processing to paper or the like. Further, the first color conversion unit 3 and the second color conversion unit 4 have a first storage device 6 and a second storage device 7 for storing input data, parameters, and the like, respectively.

The input device 1 converts a color original into a representative color and converts it into an 8-bit color separation signal of red (R), green (G), and blue (B) (hereinafter, referred to as an "RGB signal"). γ
It is sent to the correction unit 2. In the present embodiment, quantization is performed for each 8 bits. However, the present invention is not limited to this.
It may be t.

The γ correction section 2 performs level correction and non-linearity correction using a γ correction table (not shown), and sends the γ corrected RGB signals to the first color conversion section 3.

The first color conversion unit 3 converts the RGB signals into device-independent color signals L ^* a ^* b ^* , and sends them to the second color conversion unit 4.

The above L ^* a ^* b ^* indicates a color space based on an L ^* a ^* b ^* coordinate system determined by the Commission International d'Eclairage (CIE). In this L ^* a ^* b ^* coordinate system, the horizontal axis is a ^* and the vertical axis is b ^* , and the axis perpendicular to the horizontal axis a ^* and the vertical axis b ^* is L ^* . Then, saturation and hue are displayed on the a ^* b ^* coordinate plane of the coordinate system, and lightness is displayed on the L ^* axis. Therefore, in the L ^* a ^* b ^* coordinate system, brightness and chroma are expressed as being independent.

Further, in the present embodiment, for example, L ^* a ^* b ^* is used as a device-independent color signal. However, the present invention is not limited to this, and a color signal including a signal representing lightness or luminance may be used. If so, another color signal may be used. For example, YIQ or CIE L ^* u ^* v ^* . The YIQ color system is also called a YCbCr color system and a color system represented by the following matrix.

Y = 0.299R + 0.587G + 0.114B I = 0.596R−0.274G−0.322B Q = 0.221R−0.522G + 0.311B Next, the second color conversion unit 4 Color signal L ^* a ^* b ^*
For conversion into CMY, that is, cyan / magenta / yellow, which can be output by the image output device 5. Then, these CMY signals are sent to the image output device 5, and a duplicate image is completed.

Next, the first color conversion section 3 as a color conversion means which is a feature of the present invention in the above color copying machine will be described in more detail.

In the first color conversion section 3, the input R
The GB signal is converted into L ^* a ^* b ^* . In this conversion, the following neural network, that is, a neural network is used.

That is, the neural network is a circuit network that imitates the neural cell network of the brain, and is a unit (also called a neuron) corresponding to a neural cell that is a basic element of the brain.
Are connected to each other to form a network.

This neural network 10 has the structure shown in FIG.
As shown in FIG. 1, the input unit has an input layer 11, a first intermediate layer 12, a second intermediate layer 13, and an output layer 14, and for example, 3 units, 5 units, 5 units, and 3 units, respectively, so as to correspond to nerve cells. Etc. are provided.

Here, as described above, the input layer 1
The output layer 1 and the output layer 14 each include three units 15 forming a network. This is because the input data of the RGB signal is subjected to the color coordinate conversion as the output data of the L ^* a ^* b ^* color system.

Also, the intermediate layers 12 and 13
Are provided, for example, five each. The number of these units is not limited to this.

The units 15 are connected to each other with a variable connection load as shown by lines 16.

In the above neural network 10,
After passing through the neural network 10, the R, G, and B signals input at the input layer 11 are input from the output layer 14 to the quantization unit 17, where they are quantized by the quantization unit 17 to obtain the final signal. It becomes an 8-bit output signal (L ^* a ^* b ^* ).

The functions of the units 15... In the intermediate layers 12 and 13 of the neural network 10 are described with reference to FIG.
It will be described based on.

In each of the units 15, as shown in FIG. 1, the output from each unit 15 in the preceding layer is input u.
_i (i = 1 to n). Also, the weight value W shown in FIG.
_i (i = 1 to n) and the threshold value Θ ₀ are constants obtained by a method described later. Further, these constants are stored in the first storage device 6.

The above input u _i , weight value W _i , threshold value Θ
_0, the multiplication unit 21 ... and the addition unit 22, the signal u _p shown in Equation 4.

[0068]

(Equation 4)

[0069] Then, the signal u _p is converted into a signal u _out at non-linear operation unit 23. Here, the non-linear operation unit 23
Is performed as shown in Expression 5, and is expressed as shown in FIG.

[0070]

(Equation 5)

The functions of each unit 15 have been described above. The neural network 10 is formed by connecting a plurality of units 15 performing these operations.

Next, a method of optimizing the above weight values w _{1 to} w _n and the threshold value Θ ₀ will be described.

As an optimization method, a back propagation method (error back propagation learning rule) generally known in neural network theory is used. This back propagation method is practically excellent among the learning methods of the neural network 10, and is generally implemented in the neural network 10.

In the back propagation method, a desired output with respect to an input is given as a teacher signal in advance, and the difference between the actual output and the teacher signal is converted into a function, and the energy is used as the energy. The weight value and the threshold value are changed until the value converges to 0 or saturation, that is, a certain value.

The energy is a function of each weight value and threshold value, and each step of the conversion process is constituted by a continuous function, multiplication and addition. For this reason,
The energy function can be differentiated using each weight value and threshold value as variables. Then, using this differential function,
The energy function can be reduced by changing each weight value and the threshold so that the energy decreases.

The above-described general back propagation method is applied to the neural network 10 in the present embodiment.
In order to apply L ^* a ^* b ^* , first, n color patches whose values of L ^* a ^* b ^* are known are read by the input device 1, and L ^* a ^* b
^{* Get the} signal value. Then, a pair of the value of n L ^* a ^* b ^* signals and L ^* a ^* b ^* is used as a teacher signal.

That is, the data of the teacher signal used for learning of the neural network 10 is L ^* obtained by measuring n predetermined color charts with a spectrophotometer ^.
a ^* b ^* color system data. The learning data is
This is data of RGB signals obtained by inputting the predetermined color chart by the input device 1.

Here, in the present embodiment, Equation 6 is adopted when calculating the energy E as the total error.

[0079]

(Equation 6)

In Equation 6, L ^* a ^* b ^* is a teacher signal, and L ^* 'a ^* ' b ^* 'is the neural network 1
It is a predicted value by 0, that is, learning data.

Further, k _L , k _a, and k _b are coefficient values for weighting an error for each chromaticity signal. In the present embodiment, for example, k _L : k _a : k _b =
Adjust so as to be 5: 1: 1.

The above-described learning enables optimization, and the obtained weight values w _{1 to} w _n and threshold value Θ ₀ are
The information is sequentially stored in the first storage device 6.

On the other hand, the second color conversion section 4 shown in FIG. 2 can also perform signal conversion with substantially the same configuration as the first color conversion section 3 described above.

That is, in the second color conversion unit 4, when learning is performed, the teacher signal is set to CMY, and the prediction value, that is, the learning data is set to the L ^* a ^* b ^* signal, and the learning may be performed.
Energy E is

[0085]

(Equation 7)

It is sufficient to set In Equation 7, no coefficient value is set for weighting the error for each chromaticity signal. This is because in CMY, there is no effect of lightness on the L ^* a ^* b ^* color coordinates.

At this time, the weight values w _{1 to} w _n and the threshold value Θ ₀ are, as stored in the first storage device 6,
This time, it is stored in the second storage device 7.

As described above, in the color copying machine of the present embodiment, the first color converter 3 converts the color signal obtained from the input device 1 into the energy E of the three chromaticity signals by the neural network 10. At the time of calculation, coefficient values k _L , k _a, and k _b for weighting are set for errors of the three chromaticity signals L ^* a ^* b ^* for each chromaticity signal.

That is, in general, human visual characteristics are sensitive to lightness, and it is desirable that the error of the lightness component is smaller than the error of other chroma and hue components. However, in the conventional method of calculating the energy E using the neural network, the error for each chromaticity signal is simply added in an equivalent manner, and as a result, the color reproducibility is not sufficient.

However, in this embodiment, when calculating the energy E of the three chromaticity signals L ^* a ^* b ^* , the energy E of each of the three chromaticity signals L ^* a ^* b ^* is calculated. A coefficient value k _L · k _a · k _b for weighting the error is set.

Therefore, it is possible to calculate the energy E in consideration of the degree of influence of each chromaticity signal, and as a result, it is possible to provide a color copying machine capable of realizing color reproducibility according to human visual characteristics. it can.

[0092] Thus, it is possible to provide a color copier capable of obtaining a duplicate image with higher image quality than before.

Also, in the color copying machine of the present embodiment,
The first color conversion unit 3 calculates the energy E of the three chromaticity signals L ^* a ^* b ^* by the neural network 10 from the RGB color signals obtained from the input device 1. At least one of L ^* a ^* b ^* is a lightness component or a luminance component.

Therefore, the first color conversion unit 3 adjusts the weight for the error of each chromaticity signal L ^* a ^* b ^* for the lightness component or the luminance component that is sensitive to human vision. Can be.

As a result, it is possible to provide a color copying machine capable of realizing color reproducibility according to human visual characteristics.

In the above description, the luminance component is a color expression using another color system, and the three components of light intensity (luminance), hue, and color saturation are three elements of color. . And
In the color system of these three elements, similarly to the lightness L ^{* of} L ^* a ^* b ^* , the distance in the color space where the luminance is larger than the other hues and color saturations is large.

On the other hand, in the color copying machine of this embodiment,
Coefficient value k _l for weighting the brightness component or the luminance component
Is the weighting coefficient value k _a · for the other two components.
is greater than k _b, it is possible to modify the lightness component L ^* or luminance component is sensitive to the human visual more close to reality.

That is, in various chromaticity space coordinates, the lightness component L ^* and the luminance component usually have a larger spatial distance than the other components. Therefore, in order to bring the lightness component L ^* and the luminance component closer to the actual ones, as described above, the coefficient value k _l for weighting the lightness component L ^* and the luminance component is changed with respect to the other two components. should be larger than the coefficient value k _a · k _b weighting of Te.

Therefore, as described above, the coefficient value k _L · k _a
· K _b, for example, _{k L: k a: k b} = 5: 1: by adjusted to 1, to provide a color copier capable of realizing color reproduction in accordance with human visual characteristics it can.

In the color copying machine of the present embodiment,
Since the chromaticity signal is CIEL ^* a ^* b ^* , that is, L ^* a ^* b ^* defined by the International Commission on Illumination, and contains L ^* , a lightness component that is sensitive to human vision. The first color conversion unit 3 can adjust the coefficient value k _l for weighting the error of the chromaticity signal for the lightness component L ^* .

As a result, it is possible to provide a color copying machine capable of realizing color reproducibility according to human visual characteristics.

In the color copying machine of the present embodiment,
The chromaticity signal is CIEL ^* a ^* b ^* , and includes L ^* , which is a lightness component that is sensitive to human vision, and the first color converter 3 outputs the chromaticity signal of CIEL ^* a ^* b ^* . The coefficient value k _l for weighting L ^* is divided into the other two components a ^* ·
b ^* Increase adjusted than the coefficient value k _a · k _b weighting against.

Therefore, the color signal obtained from the input device 1 is
CIEL^*a^*b^*Color conversion processing to the chromaticity signal of
-When output to the image output device 5 as an image,
Of the inevitable calculation errors
Lightness component L that is sensitive to ^*More realistic
Can be modified.

As a result, it is possible to provide a color copying machine capable of realizing color reproducibility according to human visual characteristics.

[0105]

As described above, the color image processing apparatus according to the first aspect of the present invention includes color conversion means for converting a color signal obtained from an input device into three chromaticity signals by a neural network. The color conversion means weights the error of each of the three chromaticity signals for each chromaticity signal when calculating the total error of the three chromaticity signals using a neural network.

That is, in general, human visual characteristics are sensitive to lightness, and it is desirable that the error of the lightness component is smaller than the error of other chroma and hue components. However, in the conventional method of calculating the total error using the neural network, the error of each chromaticity signal is simply added, and as a result, the color reproducibility is not sufficient.

However, in the present invention, when calculating the total error of the three chromaticity signals, the error for each chromaticity signal in the three chromaticity signals is weighted. It is possible to calculate the total error in consideration of the degree of influence, and as a result, it is possible to provide a color image processing apparatus capable of realizing color reproducibility according to human visual characteristics.

According to a second aspect of the present invention, in the color image processing apparatus according to the first aspect, at least one of the chromaticity signals is a lightness component or a luminance component. is there.

Therefore, the color conversion means can adjust the weight for the error of each chromaticity signal for the lightness component or the luminance component that is sensitive to human vision.

As a result, it is possible to provide a color image processing apparatus capable of realizing color reproducibility according to human visual characteristics.

According to a third aspect of the present invention, as described above, in the color image processing apparatus according to the second aspect, the coefficient value for weighting the brightness component or the luminance component is different from that of the other components. It is larger than the coefficient value for weighting one component.

Therefore, a lightness component or a luminance component that is sensitive to human vision can be corrected to a more realistic one.

That is, in various chromaticity space coordinates, the lightness component or the luminance component usually has a larger spatial distance than the other components. Therefore, in order to bring the brightness component or the luminance component closer to the real one, as described above,
The weighting coefficient value for the lightness component or the luminance component,
It is necessary to make it larger than the weighting coefficient value for the other two components.

As a result, it is possible to provide a color image processing apparatus capable of realizing color reproducibility according to human visual characteristics.

According to a fourth aspect of the present invention, there is provided a color image processing apparatus according to the second aspect, wherein the chromaticity signal is CIEL ^* a ^* b ^* .

Therefore, the color conversion means adjusts the weight for the error of the chromaticity signal with respect to the lightness component because it includes the lightness component L ^* which is sensitive to human vision. be able to.

As a result, it is possible to provide a color image processing apparatus capable of realizing color reproducibility according to human visual characteristics.

According to a fifth aspect of the present invention, there is provided a color image processing apparatus according to the fourth aspect, wherein the CIEL ^* a ^* b ^* is weighted with respect to L ^* in the color image processing apparatus. The numerical value is larger than the coefficient value for weighting the other two components a ^* and b ^* .

Therefore, when a color signal obtained from an input device is subjected to color conversion processing to a chromaticity signal of CIEL ^* a ^* b ^* and is output to a picture output device as a color image, it is necessary to avoid the conventionally existing avoidance. Among the impossible calculation errors, a lightness component that is sensitive to human vision can be corrected to a more realistic one.

As a result, it is possible to provide a color image processing apparatus capable of realizing color reproducibility according to human visual characteristics.

[Brief description of the drawings]

FIG. 1 shows an embodiment of a color copying machine according to the present invention, and is a block diagram showing an internal structure of each unit which performs an operation using a neural network in a first color conversion unit.

FIG. 2 is a block diagram showing an overall configuration of the color copying machine.

FIG. 3 is a block diagram showing a neural network in a first color conversion unit of the color copying machine.

FIG. 4 is an explanatory diagram showing a trajectory obtained by an arithmetic expression performed by a nonlinear arithmetic unit of the color copying machine.

FIG. 5 is an overall configuration diagram showing a conventional color copying machine.

FIG. 6 is a block diagram showing an internal structure of each unit that performs an operation using a neural network in a first color conversion unit of the color copying machine.

[Explanation of symbols]

Reference Signs List 1 input device 3 first color conversion unit (color conversion means) 5 image output device 6 first storage device 7 second storage device 10 neural network 15 unit 21 multiplication unit 22 addition unit 23 non-linear operation unit E energy (total error) k _L coefficient value k _a coefficient value k _b coefficient value

Claims (5)

[Claims] 1. A color image processing apparatus for performing color conversion processing on a color signal obtained from an input device and outputting the color signal as a color image to an image output device. Color conversion means for converting the three chromaticity signals into three chromaticity signals when the neural network calculates the total error of the three chromaticity signals. A color image processing apparatus for weighting an error. 2. The color image processing apparatus according to claim 1, wherein at least one of said chromaticity signals is a lightness component or a luminance component. 3. A color image processing apparatus according to claim 2, wherein the weighting coefficient value for the brightness component or the luminance component is larger than the weighting coefficient value for the other two components. . 4. The color image processing apparatus according to claim 2, wherein said chromaticity signal is CIEL ^* a ^* b ^* . 5. The method according to claim 1, wherein a coefficient value for weighting L ^{* of} said CIEL ^* a ^* b ^* is greater than a coefficient value for weighting the other two components a ^* and b ^* . The color image processing device according to claim 4.