JP2883398B2 - Color copier - Google Patents

Description

The present invention relates to a color copying machine.

(B) Conventional technology FIG. 6 is a block diagram showing digital signal processing in a conventional general digital color copying machine.

(1) is a scanner, an R.C. obtained from a CCD sensor.
GB analog signals are converted to digital signals and output. (2) is a shading correction circuit for performing shading correction, which corrects a light distribution of the exposure light source and a variation in sensitivity of each dot of the CCD sensor. In the density conversion circuit (3), since the data read by the scanner is reflected light data, the data is converted into density data.

(4) is a black extraction and under color removal circuit. Here, the reproducibility in the shadow portion is improved by adding the black toner to the three-color toner to supplement the density in the high density range. Also, by replacing each color toner with black toner, the consumption of each color toner can be reduced. Black extraction under color removal circuit (4)
And Y (yellow), M (magenta) and C (cyan)
A signal obtained by adding K (black) to the color is input to the color / gradation correction circuit (5). Here, since the spectral characteristics of each color toner are not the ideal type, a color correction technique such as a masking process is used. (6) is a sharpness correction circuit, which enhances image data at the edge portion of a character / line image to improve image quality. (7) receives the signal from the color / gradation correction unit,
This is a multi-level signal circuit for converting into a binary or multi-level signal according to the printer. (8) is a printer device such as a laser printer.

The color / gradation correction circuit (5) as a color correction means in the above-mentioned conventional digital color copying machine often uses a linear masking method which requires small hardware. In a digital color copying machine, Y (yellow), M (magenta), and C (color) are input based on R (red), G (green), and B (blue) signals input from a scanner (1).
(Cyan) performs the same color reproduction as the original with each toner. That is, a change is made from R, G, and B, which are additive signals, to Y, M, and C, which are subtractive signals.

Now, for a document having spectral reflection characteristics divided into R, G, and B regions as shown in FIG. 7, the reflection intensity of each spectrum is defined as (R ₀ , G ₀ , B ₀ ). A method of reproducing this so as to have the same spectrum in a subtractive color system will be considered. Since subtractive color system is a method of gradually Hesi therefrom illuminates the white light, to produce a similar spectral distribution and FIG. 7 is a light intensity of the white light of the shaded portion of Figure 7 from R _W Energy needs to be reduced, or absorbed. B area B
In the same manner, it is sufficient to reproduce the toner Y with G, the G with G having M with G, and the R with C having R.

This relationship is calculated from the reflectances R, G, and B based on the toner densities DY, D.
The conversion to M and DC has the relationship shown in equation (1).

DY = -logB DM = -logG DC = -logR (1) However, the above applies to the idealized model. In fact, for example, the C dye absorbs not only red but also green and blue light. There is an unnecessary absorption component that does. Since these unnecessary absorption components cause color turbidity, DY, DM, and DC obtained by Expression (1) are normally corrected by the following expression by a linear masking method as shown below.

DC ', DM', and DY 'are the toner amounts to be obtained.

In practice, a large number of color patches are usually output from a printer using the above formula, and coefficients are often estimated by the least squares method so that the colors of the original and the printed matter match.

On the other hand, in the above-described conventional digital color copying machine and the like, the color balance is adjusted by operating a panel on the copying machine. For example, in the color laser copier 1 (CLC1) manufactured by Canon Inc., in the color balance adjustment mode, Y (yellow), M (magenta),
(Cyan) and K (black) are instructed from the panel to increase or decrease the amount of each toner.

(C) Problems to be Solved by the Invention However, the above-described linear masking method cannot sufficiently cope with the non-linearity due to the addition law of the color density component and the failure of the proportional law in the superposition of the three color toners. There is room for improvement in color reproducibility.

According to the above-described color balance adjustment method, for example, when it is desired to make a copy original a little more reddish, a color material Y (yellow) and a green light (500n) that absorb blue light (400n to 500n) are used. It is necessary to increase the amount of color material M (magenta) that absorbs 〜600 n), but it is rare that a general copying machine user has knowledge of these colors.

Therefore, when color balance adjustment is performed by a person who does not have knowledge about colors, it is necessary to consider which color material to increase or decrease while reading the manual.

Also, after deciding which color material to increase or decrease, it is difficult to judge how much to increase or decrease from the scale on the operation panel, so it is necessary to copy many times to obtain the desired color balance There was no way.

SUMMARY OF THE INVENTION The present invention has been made to solve the above-described conventional problems, and provides a color copying machine in which color adjustment is easy.

(D) Means for Solving the Problems The present invention includes a voice input instructing means for inputting and instructing a color balance by voice, and a color setting means for determining how much the toner of each color should be increased or decreased. An instruction output from the voice input instruction means, at least an input layer,
An input is made to an input layer of the color setting means composed of a neural network having an intermediate layer and an output layer, and a toner density for each color is determined based on an output from the output layer of the color setting means.

(E) Function In the present invention, when performing color balance adjustment,
By inputting what color (for example, to increase reddishness) and how much (for example, to increase reddishness a little), the color balance is copied as expected by the user. A manuscript can be created.

(F) Example Hereinafter, an example of the present invention will be described with reference to the drawings.

 The same parts as those of the conventional example are denoted by the same reference numerals.

FIG. 1 is a schematic block diagram showing an embodiment on which the present invention is based.

(1) is a scanner, an R.C. obtained from a CCD sensor.
GB analog signals are converted to digital signals and output. (2) is a shading correction circuit for performing shading correction, which corrects a light distribution of the exposure light source and a variation in sensitivity of each dot of the CCD sensor. In the density conversion circuit (3), since the data read by the scanner is reflected light data, it is converted into density data as in the above equation (1).

A color / gradation correction circuit (50) determines the amounts of Y, M, and C toners using a neural network for color correction.

(6) is a sharpener correction circuit, which enhances image data at the edge portion of a character / line image to improve image quality.
(7) is a multi-level signal circuit which receives a signal from the color / gradation correcting section and converts it into a binary or multi-level signal corresponding to the printer. (8) is a printer device such as a laser printer.

Reference numeral (20) denotes a host device such as a personal computer for providing a teacher signal to be described later to the neural network of the color / gradation correction circuit (50). The color / correction circuit (50) transmits data to and from the host device. An interface for transmitting and receiving is provided.

The color / gradation correction circuit (50) performs correction in consideration of the increase / decrease of each color signal, and sends data to the sharpness correction circuit (6). Subsequent processing is the same as in FIG.

Next, a neural network used in the present invention will be described with reference to FIG. FIG. 2 is a schematic diagram showing a neural network used in this embodiment.

The neural network in this embodiment has a three-layer structure, and there is no connection between units in the same layer. Each unit is connected to each of the units in another layer via a weight coefficient (synaptic weight). Density conversion circuit (3) Density-converted density values DY, DM, DC
Are supplied as input signals to the respective units, and the toner amounts DY ', DM', and DC 'are output from the respective units in the output layer.

 The learning method uses back propagation.

In the back propagation, a sigmoid type function y = ^1/1 + e- ^x as shown in FIG. 3 is considered as an input / output function of the neuron.

The total input to a certain neuron J is the output of the neuron y _i that is synapse-coupled to the neuron J and the synapse weight.
The ones that have been multiply-accumulate in the W _ji.

The output of the input to the neuron J calculated by the above equation is defined as the following equation.

y _j = ^1/1 + e ^-xj (3) Using the above two equations, the signal given to the input layer is sequentially calculated according to the synaptic weight of the network from the input layer to the intermediate layer to the output layer. The calculation result of the output layer is compared with the teacher signal. Specifically, assuming that the i-th output layer is y _i and the teacher signal is d _i , the degree of deviation of the teacher signal is expressed by the following equation as an error E.

E = 1 / 2Σ (d _i −y _i ) ² (4) Backpropagation modifies the synapse wit of the network so as to minimize this error E.

The learning signal δ is represented by the following equation for the correction from the output layer to the hidden layer.

δ _i = (d _i −y _i ) f (u _i ) (5) where u _i is the i-th neuron state shown in equation (2), and f is a sigmoid expressed by equation (3). It is a type function.

Further, the learning signal δ of each intermediate layer is obtained by the following equation considering the neuron k in the layer one stage higher than the neuron j.

Then, based on the equations (5) and (6), the synapse weight is obtained by the following rule.

ΔW _ij (n + 1) = η (δ _i y _i ) + αΔW _ij (n) (7) where η is a learning constant and α is a stabilization constant.

Expression (3), which is an input / output function, becomes f '= f (1-f), and the operations of expressions (5) and (6) can be easily performed.

In the learning using the backpropagation described above, specifically, a large number of color originals are output from a printer, and the color misregistration with the printed matter is observed. The judgment is made by a human using a measuring instrument or the like. If it is judged that there is a color shift, a personal computer of the host (20) inputs a density value at which no color shift is expected to occur. This density value is used as a teacher signal, and Pacoson reads out the current synapse weight on the neural network by this input, as described in (5) and (6) above.
An arithmetic operation is performed using equation (7) to determine a new synapse weight, and the value is written to the color / tone correction circuit (50). By repeating such an operation and changing the document, the color shift gradually disappears and the network converges.

By the way, the host device (20) composed of Pacoson and the like shown in FIG. 1 is used at the time of learning, and it is needless to say that it becomes unnecessary after the learning is completed.

In the embodiment of FIG. 1, a color copier for printing using three color toners has been described as an example.
The same applies to color reproduction using four-color toners including.

 Further, the multi-level conversion circuit (7) may be configured by binarization.

Further, the more the number of intermediate layers in the neural network, the finer the color reproduction. However, in that case, hardware becomes large.

FIG. 4 is a schematic block diagram showing an embodiment of the present invention.

The same parts as those in FIGS. 1 and 6 are denoted by the same reference numerals, and description thereof is omitted.

Reference numeral (10) denotes a voice input unit for inputting a color balance adjustment instruction by voice. Of course, it can be used for settings other than the color balance adjustment instruction. (11) is a speech recognition device, which receives a speech signal received from the speech input of the speech input unit (10), analyzes it to determine what words are inputted, and outputs the result.

(12) is a neural network unit as color setting means, which receives a signal from the speech recognition device (11) and
(Yellow), M (magenta), C (cyan), and K (black) are determined and output, and output. The output result is input to the color / gradation correction circuit (50). The color / gradation correction circuit (50) performs correction in consideration of the increase / decrease of each color signal, and sends data to the sharpness correction circuit (6). Subsequent processing is the same as in FIG.

By the way, the learning of the color balance adjustment using the neural network is performed as follows.

First, a certain color original is copied, and for example, when it is desired to slightly increase the red tint of the copied original, "slightly darker red" is input from the voice input unit (10). The input speech is converted by a speech recognition device (11) into information representing hue (red, blue, orange, etc.), information representing saturation (light, dark, dark) and information representing lightness (bright, dark). Furthermore, they are classified into information that modifies them (extremely, considerably, somewhat more or less), and the information is output to the neural network (12).

FIG. 5 shows a correspondence diagram between the above information and the input layer of the neural network corresponding to Y (yellow). In this example, the neurons corresponding to “a little more”, “red”, and “dark” emit light. On the other hand, in the output layer, the amount
In the example of the figure, the rate of increase / decrease with respect to the current toner density is shown, and as a result of calculating the input signal, the rate corresponding to the neuron that responds most in the output layer is set as the output result.

In this way, for the input "A bit more red,"
Y (yellow), M (magenta), C (cyan), K
(Black) The output of the neural network (12) corresponding to each color is input to the color / gradation correction circuit (50),
In consideration of the result, the toner densities of Y (yellow), M (magenta), C (cyan), and K (black) are determined. Obtain a copy.

If the desired color balance is obtained from the copy original obtained in this way, the weight of each synapse of the neural network of each color is strengthened. For example, when it is determined that the Y component is too strong, the neural network for Y (yellow) is used. Learn the correct answer for the remaining M
For (magenta), C (cyan) and K (black), the weight of each synapse is strengthened.

The learning described above is repeated for one document under the same input conditions many times. Further, learning is repeatedly performed by changing the input conditions for the same original. Furthermore, we change a manuscript and continue learning,
Regardless of the input condition and the input document, the learning is terminated when a general user can obtain a satisfactory copy.

The neural network (12), which has been trained as described above, is incorporated into a copier and provided to the user. The user inputs a voice from the voice input device (10) so as to obtain a desired color balance, and the learning is performed. A copy original can be obtained in the same manner as at the time.

By the way, it is needless to say that the learning is performed until the copying machine becomes a product, and does not cause the user to perform the learning. However, it is possible to further learn the neural network after learning, and if it is possible to incorporate such an operation procedure into a copying machine,
It is also possible to make the user learn.

In the above-described embodiment, the voice input unit (10) is used as a means for instructing color balance. However, the present invention is not limited to this. For example, a key that displays the type of color and its intensity on the operation panel is used. And the like.

Further, by providing a means for recognizing an object (eg, face, body) in a document, the present invention is not limited to the adjustment of the overall color balance, but also performs a partial color adjustment (for example, making the face paler). Becomes possible.

(G) Advantages of the Invention As described above, according to the present invention, when color balance is adjusted, a person who has knowledge about colors can perform Y (yellow), M (magenta), Although the amount of toner of C (cyan) and K (black) has been increased or decreased, even a person who does not have special knowledge about color can make adjustments by voice and the operation is simplified. .

[Brief description of the drawings]

FIG. 1 is a block diagram showing an embodiment on which the present invention is based;
FIG. 2 is a schematic diagram showing an example of a neural network,
FIG. 3 is a characteristic diagram showing an output function of a neuron. FIG. 4 is a block diagram showing an embodiment of the present invention, and FIG.
It is a corresponding figure of the neural network corresponding to (yellow). FIG. 6 is a block diagram showing the configuration of a conventional color copying machine;
FIG. 7 is a schematic diagram showing the spectral reflection characteristics. DESCRIPTION OF SYMBOLS 1 ... Scanner, 2 ... Shading correction circuit, 3 ... Density conversion circuit, 4 ... Black extraction / under color removal circuit, 5, 50 ... Color / gradation correction circuit, 6 ... Sharpness correction circuit, 7 ... Multi-value conversion circuit, 8... Printer, 10... Voice input unit, 11... Voice recognition device, 12.

Claims (1)

(57) [Claims] 1. A color copying system in which an original is separated into colors, read, color conversion and color correction are performed, an electrostatic latent image for each color is formed on a photoreceptor, and these are sequentially transferred onto a sheet for reproduction. A voice input instructing means for inputting and instructing a color balance by voice; and a color setting means for determining how much the toner of each color should be increased or decreased, and an instruction from the voice input instructing means. An output is input to an input layer of the color setting means comprising a neural network having at least an input layer, an intermediate layer, and an output layer, and a toner density for each color is determined based on an output from the output layer of the color setting means. A color copier characterized by: