JP2855938B2 - Color imaging equipment - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a color image apparatus for converting a color image into an electric signal in a color facsimile or a color copying machine.

[0002]

2. Description of the Related Art For example, FIG.
FIG. 1 is a configuration diagram of a signal processing unit of a color copying machine disclosed in Japanese Patent Application Laid-Open Publication No. HEI 10-202, FIG. Reference numeral 1 denotes a conversion unit that receives y (yellow), m (magenta), and c (cyan) as input color signals, corrects color and gradation using a neural network, and determines toner density. Reference numeral 9 denotes a host computer that supplies learning data to the neural network. FIG. 6 is a configuration diagram showing a neural network of the conversion unit 1 of the signal processing unit of the color copying machine. The neural network has a three-layer structure. There is no connection between neurons in the same layer, and each neuron is connected to all neurons in other layers via weights. Here, 11i is an input layer neuron, 11h is an intermediate layer neuron, and 11o is an output layer neuron.

Next, the operation will be described with reference to FIG. The input color signal (ymc) and the output color signal (YM
In order to configure the network of the conversion unit 1 (FIG. 6) so as to be suitable for C), each weight of the network and learning data (hereinafter, s is added) representing the relationship between the input color signal and the output color signal are used. The optimization of the offset (referred to as learning) is performed first. Each weight and offset are initialized prior to learning. (Step A1). Next, the input color signal (ymc) _s of the learning data is input to the conversion unit 1. (Step A2). The conversion unit 1 performs an operation using the current weights and offsets, and obtains an output.
Conventionally, the learning of the neural network by the conversion unit 1 has been performed using back propagation. Neuron 11
Can be represented by symbols as shown in FIG. Where In is n
The third input, O, represents the output. Each input is given a weight Wn representing the strength of the connection. Each neuron is given an offset value θ, and the output O
Is

[0004]

(Equation 1)

[0005] Here, the function f is generally a sigmoid function f (x) = 1 / (1 + exp (-x))
Is used. (Step A3) Output data (YMC) _{s of the} learning data and output data (YM
The error of C) is obtained (step A4). Each weight is set so that the square error is reduced based on the error obtained in step A4.
Change the offset slightly. The change is made, for example, by the fastest descent method (step A5). Next, the learning data ((ym
c) Update _s and (YMC) _s ) (Step A)
6). At this time, it is determined whether all learning data has been completed,
If not completed, steps A2 to A6 are repeated (step A7). When the process is completed for all the learning data, it is determined whether or not the total error of the output color signal for each learning data is equal to or less than a certain value. If the value is larger than the predetermined value, the learning data is returned to the original value, and Step A2 is performed.
To step A7 are repeated (step A8). If the sum of the errors of the output color signals with respect to the learning data is equal to or smaller than a certain value, it is assumed that the network optimization (learning) has been completed.

[0006]

The conversion unit 1 of the conventional color copying machine learns by back propagation and is constituted by a neural network having the above-described structure. There was a problem that the number of neurons in the hidden layer had to be increased, and the hardware became large.

The reason will be described below with reference to FIG. Focusing on one of the hidden neurons 11h, 3
It is connected to two output layer neurons 11o. Therefore,
The output of the intermediate layer neuron 11h is input to three output layer neurons 11o via the weight of each connection. In the case of (a), the output neuron 11o of Y is promoted, and the neuron 11o of M is promoted.
It acts in a suppressive manner on the C output layer neuron 11o.
In the case of (b), it acts on the Y output layer neuron 11o in an accelerated manner and suppresses on the M and C output layer neurons 11o. Also, there are a large number of weights of the connection, from small to large.

When the weight of the connection is determined, one intermediate layer neuron 11h outputs the output layer neuron 11o.
Is fixed. Since the weight of the connection is determined by the relationship with the learning data of the three output layer neurons 11o that are connected, the weight of the connection is an intermediate value of the three neurons. Therefore, in order to improve the accuracy of color reproduction, it is necessary to increase the number of hidden neurons 11h and prepare a large number of hidden neurons 11h with different degrees of coupling to the three output neurons 11o.

Three input layer neurons 11i, nine intermediate layer neurons 11h, and three output layer neurons 11o
When an experiment was performed using a neural network composed of the above, the average error between the converted value of the input signal and the expected output value was 5%, and the maximum error was 10%.

As described above, in the color / tone correction using the conventional neural network, the number of neurons in the intermediate layer must be increased in order to improve the color reproduction accuracy, and the hardware becomes large. was there.

SUMMARY OF THE INVENTION The present invention has been made to solve the above problems, and has as its object to obtain a color image apparatus using a neural network having high color reproduction accuracy using a small number of hidden neurons.

[0012]

A color image apparatus according to the present invention converts a color image signal into a neural network.
Correction means for correcting the color by using the neural network
Change the weight and offset of the work
Correction of color image signal by learning means for optimization
Input signal in a neural network
And a conversion unit for converting the input signal into an output signal.
A storage unit for storing learning data indicating a relationship between the output signals
And learning data of the storage unit and an output signal of the conversion unit.
And a neural network of the conversion unit.
Network weight so that the error of the subtractor is reduced.
And a weight changing unit .

[0013] The color correcting and correcting means for correcting the color of the color image signal by the neural network, a color image signal and a learning means for performing optimization by changing the weights and offsets Men u neural network Network In the image apparatus, the configuration of the neural network of the conversion unit forming the main part of the correction means is divided into n (n ≧ 3) blocks for the output color signal , and each neuron of the intermediate layer is only one output layer neuron. This is to correct the color image signal by combining them.

[0014]

In the color image apparatus according to the present invention, since the intermediate layer neurons of the neural network constituted by the conversion unit 1 are connected to only one output layer neuron, the connection between the final output data and the intermediate layer neurons is strong. That is, the number of times of learning for obtaining a given input color signal and an expected output color signal corresponding thereto is reduced, and the number of constituent neurons can be reduced.

[0015]

[Embodiment 1] An embodiment of the present invention will be described below with reference to the drawings. FIG. 1 is a diagram showing a configuration of a signal processing unit of the present invention. In FIG. 1, reference numeral 1 denotes a conversion unit that converts an input signal into an output signal; 2 denotes a weight change unit that changes the weight of the conversion unit 1 based on an error between a color signal in a storage unit 3 and an output color signal; A storage unit 4 for storing learning data representing a relationship between the color signal and the output color signal; a selector 4 for switching an input signal between a color signal from the storage unit 3 and a color image signal from a sensor (not shown); Is a subtractor for calculating an error between the color signal of the storage unit 3 and the output color signal.

FIG. 2 is a diagram showing a neural network of the converter 1. In FIG. 2, 11i is an input layer neuron, 11h is an intermediate layer neuron, and 11o is an output layer neuron. Image data r, g, and b are input to each of the three input layer neurons 11i, and the output is input to all the intermediate layer neurons 11h. One output layer neuron 11o does not include all the neurons in the intermediate layer, but three predetermined intermediate layer neurons 11h.
Is input. Each output layer neuron 11o
The signal corresponding to R, G, B is output. As described above, the conversion unit 1 is configured by a neural network including a combination of neurons.

The operation of the present invention will be described below. Learning of the neural network shown in FIG. 2 is performed by back propagation. That is, the selector 4 is switched so that the learning data of the input data of the storage unit 3 is input to the conversion unit 1, and an error between the data converted by the conversion unit 1 and the learning data of the output data of the storage unit 3 is subtracted. Each weight and offset are slightly changed by the fastest descent method so that the square error is reduced based on the error. On this occasion,
As shown by the broken line and the chain line in FIG. 2, the structure of the neural network is divided into three blocks for the output color signal. Therefore, the intermediate layer neuron 11h is connected to only one output layer neuron 11o. The connection between the final output data and the intermediate layer neuron 11h becomes stronger, and the connection between the final output data and the other output layer neuron 11o becomes weaker. This is suitable for data that is theoretically decomposed into three primary colors, such as image data. for that reason,
The number of learning iterations can be reduced, and the number of constituent neurons can be reduced.

The relationship between the input data and the output data of the network thus optimized is, for example, as shown in FIG. Here, ● indicates a set (x, y) of the input color signal and the output color signal of the learning data. As shown in FIG.
The feature is that a curve faithful to the learning data is obtained as an optimal solution.

As a result of the experiment, in the optimized network, the average error between the signal value obtained by inputting the input color signal (rgb) of the learning data to the conversion unit 1 and the output color signal (RGB) of the learning data is: 3%, maximum error is 5
%.

To convert the input color signal from the sensor, the selector 4 is switched to the input color signal from the sensor, and the conversion is performed using the optimized network weight and offset.

Embodiment 2 FIG. In the first embodiment, the conversion unit 1 and the weight change unit 2 have a configuration capable of learning. However, as shown in FIG. 4, the conversion unit 1 is configured only with the ROM 8, and learning is performed by an external computer. Then, the optimized relationship between the input color signal and the output color signal of the network may be stored in the form of a table in the ROM 8, and the color may be corrected.

[0022]

As described above, according to the present invention, the color correction is performed in the form of a neural network having a structure in which the output layer neuron 11o is divided into n blocks, and the intermediate layer neuron of the neural network has one output. Input color signals and output color signals that have a complicated relationship because they are connected only to layer neurons are realized by a neural network with the same number of constituent neurons and further reduce the error of each color, color reproduction There is an effect that a highly accurate color image device can be obtained.

[Brief description of the drawings]

FIG. 1 is a configuration diagram of a signal processing unit of a color image device according to an embodiment of the present invention.

FIG. 2 is a diagram showing a neural network of a conversion unit 1 of the color image device according to one embodiment of the present invention.

FIG. 3 is a diagram showing a relationship between input data and output data of an optimized network of the color image device according to one embodiment of the present invention;

FIG. 4 is a configuration diagram of a signal processing unit of an image device according to another embodiment of the present invention.

FIG. 5 is a configuration diagram of a signal processing unit of a conventional color copying machine.

FIG. 6 is a configuration diagram of a neural network of a conversion unit 1 of a signal processing unit of a conventional color copying machine.

FIG. 7 is a diagram of symbols showing neurons.

FIG. 8 is a flowchart showing the operation of color correction of a conventional color copying machine.

FIG. 9 is a diagram showing a connection between an intermediate layer neuron and an output layer neuron of a signal processing unit of a conventional color copying machine.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Conversion part 2 Weight change part 3 Storage part 5 Subtractor 11h Intermediate layer neuron 11o Output layer neuron

Claims (2)

(57) [Claims] And correcting means for correcting the color by 1. A neural network a color image signal, e Bei and learning means for performing optimization by changing the weights and offset of the upper Symbol neural network network, correcting the color image signal in a color image apparatus, an input signal neural that
A conversion unit for converting to an output signal by a network ,
Stores learning data indicating the relationship between the input signal and the output signal
A storage unit that performs learning data of the storage unit and
A subtractor for obtaining an error from the output signal;
Reduce the error of the subtractor
And a weight changing unit for changing the color image. 2. A correcting means for correcting the color by the color image signal neural network, e Bei and learning means for performing optimization by changing the weights and offsets Men u neural network network, the color image signal In the color image apparatus for correction, the configuration of the neural network of the conversion unit forming the main part of the correction means is divided into n (n ≧ 3) blocks for the output color signal.
Wherein each neuron is coupled to only one output layer neuron to correct the color image signal.