JPH0738745A - Image forming device setting picture quality by neural network - Google Patents

Abstract

PURPOSE:To easily and automatically set picture quality corresponding to the choice of each user with simple configuration. CONSTITUTION:Based on an ID password, a neuro computer 41 performing learning is selected by a switch 61. The selected neuro computer 41 performs learning with 15 pieces of histogram data from histogram counters 52...52 as input data and with picture quality data from an operation part 3 as teacher data. At the time of copying after learning, the neuro computer 41 designated by the ID password calculates the picture quality data suitable for the inputted histogram data according to the contents of learning. An internal parameter part 59 outputs an internal parameter corresponding to the picture quality data outputted from the neuro computer 41. A process control part 63 outputs a process control value corresponding to the internal parameter, and a copy process part 6 is controlled by this process control value. As a result, copying is performed with the picture quality suitable for an original image.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention uses a neural network that learns an image of an original document and the image quality associated with each image by a neural network, and forms an image with the image quality set according to the learning content. The present invention relates to an image forming apparatus that sets image quality.

[0002]

2. Description of the Related Art Generally, a copying machine is adapted to perform copying in a standard mode in which copying conditions such as copying image quality are set to standard values. However, when the copying is performed in the standard mode, the optimum setting of the copying condition varies depending on the image of the document and the individual user's preference for the image, so that the copying cannot always be performed with a satisfactory image quality. Therefore, the conventional copying machine described below is configured so that the setting of the copying condition can be changed.

For example, in a copying machine having a function called a job memory or the like, the copying condition is registered in advance by the function, and the registered copying condition is called when copying is required. ing.

Further, Japanese Patent Laid-Open No. 15866/1993 discloses that
A copier has been disclosed that informs the necessity of an automatic document feeder according to a copy history. This copier specifically
Store copy history such as document size and number of copies,
Based on the copy history, the frequency of continuous document replacement within a predetermined time is calculated, and the necessity of the automatic document feeder is determined from the frequency.

Further, Japanese Patent Laid-Open No. 4-273650 discloses a copying machine which compares image data of an original with image data of a copied image of the original and adjusts the copy image quality based on the comparison result. There is.

[0006]

However, in the copying machine having the above-mentioned job memory function, the panel operation is required each time the copying condition is registered or called, so that the operation tends to be complicated. . In addition, Japanese Patent Laid-Open No. 3-158
In the copying machine disclosed in Japanese Patent No. 66 and Japanese Patent Laid-Open No. 4-273650, the parameters are reduced so that the function for setting the copying conditions is simple. Therefore, the relationship between the image information of the document and the user's favorite color tone, density, etc., is too large for the parameters,
It cannot be stored as a copy condition.

In order to make this possible, a method of storing the image information of the original document and the copy condition in a one-to-one correspondence and automatically setting the copy condition from the image information read from the original document can be considered. . However, such a method requires a large capacity memory as the number of stored data increases. In addition, in order to accurately perform the approximation when the read image information is different from the stored image information, innumerable stored data are required. Further, even if a large-capacity memory is prepared, there is a problem that data cannot be retrieved instantaneously unless the data structure for storage is devised.

The present invention has been made in view of the above circumstances with respect to the setting of the image quality in which the user's preference is remarkably reflected among the copy conditions, and the relationship between the image information of the original document and the image quality condition is neural-coded. It is an object of the present invention to provide a copying machine in which image quality conditions can be easily set with a simple structure by learning through a network.

[0009]

In order to solve the above-mentioned problems, an image forming apparatus for setting the image quality by a neural network of the present invention, in order to solve the above-mentioned problems, an image information reading means for reading the image information of a document image, and this image information reading. An image forming means for forming a copy image on a transfer material based on image information read by the means, an image quality condition input means for inputting image quality conditions of the copy image, and an image quality condition input from this image quality condition input means. Control means for controlling the image forming means based on the above, and image data read by the image information reading means as input data, and the image quality condition input from the image quality condition inputting means as teacher data. Learn the relationship with the data by associating it with the identification number and learning the identification number specified when copying the original image. And a quality condition determining means having a neuro-computer for calculating the image quality condition suitable for the image information read by the image information reading means in accordance with contents.

Further, the image forming apparatus described above, the converting means for converting the image quality condition input from the image quality condition inputting means into a parameter suitable for controlling the image formation, and the image forming based on the parameter from the converting means. An image forming apparatus comprising a control value setting means for setting a control value for controlling the means, wherein the control means controls the image forming means based on the control value set by the control value setting means, Further, the image information read by the image information reading means is used as input data, the parameter from the converting means is used as teacher data, and the relationship between the input data and the teacher data is learned for each identification number, and the original image is copied. When the identification number is specified when performing the image reading, the image read by the image information reading means is read according to the learning content about the identification number. And a parameter determining means having a neuro-computer for calculating the parameters for the information.

Further, in the above image forming apparatus, the image quality condition determining means or the parameter determining means has the following respective features.

(1) A processing section for performing learning and calculation according to the learning content, a plurality of learning content storage sections for storing the learning content, provided for each recognition number, and a learning content storage section for a designated recognition number. And a connection switching means for connecting to the processing unit.

(2) The learning contents are stored in an external storage medium.

(3) The learning content for each recognition number is stored in each of the plurality of external storage media.

(4) A data storage unit for storing, as learning data, a pair of input data and teacher data used for learning is provided, and learning is performed based on the learning data stored in this data storage unit.

(5) Learning is carried out after a certain amount of learning data is accumulated in the data accumulating section.

(6) Only the last input learning data for the same document is stored in the data storage section.

(7) If it is determined that the input learning data is similar to the learning data already stored in the data storage unit, the learning data is excluded from the target to be stored in the data storage unit.

(8) A basic learning calculation unit for learning the relationship between the standard image information and the image quality condition or control value in advance, and the output of this basic learning calculation unit for calculating the image quality condition according to the learning content. The subtraction unit that calculates the difference between the value and the teacher data, and the output of the subtraction unit is used as the difference teacher data to learn the relationship between the difference data and the input teacher data. A difference learning calculation unit that calculates a difference from the output value of the calculation unit and an addition unit that calculates the sum of the output value of the difference learning calculation unit and the output value of the basic learning calculation unit are provided.

(9) The learning content of the difference learning calculation section is deleted for each recognition number.

Further, the image forming apparatus having each of the above features shares the contents learned by the image quality condition determining means or the parameter determining means with an external device having the same function as the image quality condition determining means or the parameter determining means. Another feature is that they are interconnected as much as possible.

[0022]

In the image forming apparatus which is the most basic of the present invention, the relationship between the image information of the original image and the image quality condition is learned by the neurocomputer prior to copying. At this time, when the image information is read by the image information reading unit, the neurocomputer included in the image quality condition setting unit learns the relationship between the image information and the image quality condition input from the image quality condition input unit for each identification number. As the image information, for example, in the case of a full-color original, the distribution state of each color of red, green, and blue, the density of each color, and the like can be mentioned.

When making a copy, the image information of the original image is read by the image information reading means. At this time, the neurocomputer performs an operation according to the relationship between the image information learned about the specified identification code and the image quality condition, and sets the image quality condition suitable for the read image information. Then, the control unit controls the image forming unit on the basis of the image quality condition, so that copying is performed.

The above-mentioned image quality forming apparatus operates as follows when controlling the image forming means by using the parameter obtained by converting the image quality condition into the internal parameter. First, the relationship between the image information of the document image and the parameters is learned by the neurocomputer prior to copying. At this time, the neurocomputer included in the parameter setting unit learns the relationship between the image information read by the image information reading unit and the parameter converted from the image quality condition by the conversion unit for each identification number.

When making a copy, the image information of the original image is read by the image information reading means. At this time, the neurocomputer performs calculation according to the relationship between the image information learned about the specified identification number and the parameter, and sets the parameter suitable for the read image information. Then, the control value setting means sets the control value by the parameter. Then, the control unit controls the image forming unit based on the control value, so that copying is performed.

As described above, the image forming apparatus learns the relationship between the image information and the image quality condition or the relationship between the image information and the parameter, and sets the image quality condition according to the learning content. Therefore, according to this image forming apparatus, it suffices to store the coupling weight of the neurocomputer as the learning content, and it is not necessary to provide a large-capacity memory as in the configuration in which the above relationship is stored in the memory or the like. Moreover, if the user uses a unique identification number by performing learning for each identification number and calculation according to the content of the learning, it is possible to easily obtain a copied image having an image quality according to the preference of each user.

The above-mentioned image forming apparatus operates as follows because the image quality condition determining means or the parameter determining means individually have each of the features (1) to (9).

(1) During learning and when setting the image quality, the connection switching means connects the learning content storage section of the designated identification number to the processing section. This not only saves the learning content for each recognition number, but also
Only one processing unit is required.

(2) At the end of learning, the learning content is stored in the external storage medium. As a result, by mounting the external storage medium on another image forming apparatus, the learning contents described above can be used by the image forming apparatus.

(3) At the end of learning, the learning content is stored in the external storage medium corresponding to the designated identification number.
This facilitates management of the learning content for each recognition number when sharing the learning content with other image forming apparatuses.

(4) Prior to learning, a certain amount of learning data is accumulated in the data accumulation section. Then, the learning data accumulated in the data accumulating unit is given to the neurocomputer and used for learning. This allows
Learning is performed based on a large amount of learning data, and more accurate image quality setting can be performed using the learning content.

(5) When a certain amount of learning data, such as 100 sets, is accumulated in the data accumulating section, the neurocomputer performs learning using the learning data.
As a result, insufficient learning will not be performed in the state where the amount of learning data accumulated is small.

(6) When learning data is obtained for the same original, learning data is sequentially input to the data storage unit, but only the last input learning data is stored. As a result, a plurality of learning data will not be adopted for the same original document, and the inconvenience that the learning of the neurocomputer will not converge can be prevented.

(7) It is determined whether the input learning data is similar to the learning data already stored in the data storage unit. At this time, for example, the difference between the values of the elements of both learning data is calculated, and if the sum of the calculation results is smaller than a predetermined value, it is determined that both learning data are similar. It Then, when it is determined that the two learning data are similar, the input learning data is excluded from the target to be stored in the data storage unit. As a result, similar learning data is not stored in the data storage unit, and learning of different patterns is performed more.

(8) During learning by the difference learning calculation unit, the basic learning calculation unit calculates and outputs a basic image quality condition based on the input image information. The subtraction section calculates the difference between this output value and the given teacher data. The difference learning calculation unit learns the relationship between the input image information and the difference teacher data by using this difference as the difference teacher data.

At the time of setting the image quality, the difference learning calculation unit calculates according to the learning content based on the input image information.
On the other hand, the basic learning calculation unit calculates and outputs a basic image quality condition based on the input image information as in the above learning. Then, in the addition unit, the output value of the difference learning calculation unit and the output value of the basic learning setting unit are added.

The value thus obtained is the basic image quality condition plus the value learned according to the user's preference. Therefore, basic image quality setting can be performed even when the number of learning is small. With such a configuration, since the basic image quality is ensured, the image quality to be set gradually becomes more preferable to the user's preference as a result of repeated learning.

(9) For example, when changing the learning content,
It is necessary to erase the learning contents. In this case, according to the configuration in which the learning content of the difference learning calculation unit is deleted for each recognition number, only the learning content corresponding to the specific identification number can be deleted. When the learning content is changed or unnecessary learning content occurs for a specific identification number, the learning content can be easily erased.

Further, since the image forming apparatuses having the above-mentioned respective features are connected to each other so that the learning contents can be shared with the external equipment, the same is true in both the image forming apparatus and the external equipment. You can use the learning content. In particular, if an image forming apparatus having the same function as this image forming apparatus is prepared as an external device, for example, the learning content stored in the image forming apparatus serving as the parent machine is called to the image forming apparatus serving as the slave machine. It is possible to use it and to return the learning contents modified by the child machine to the parent machine.

[0040]

<Embodiment 1> An embodiment in which the present invention is applied to a full-color copying machine will be described below with reference to FIGS. 1 to 27.

[Structure and Basic Operation of Copying Machine] As shown in FIG. 2, the copying machine according to the present embodiment has a transparent document placing table 2 and an operating section 3 which will be described in detail later on a top surface of a main body 1. It is set up. A reading optical system 4 and a reproducing optical system 5 are arranged below the document table 2. Below the reading optical system 4, a copying process unit 6 as an image forming unit is arranged.

The reading optical system 4 has an exposure lamp 7, mirrors 8 ... 8, a lens unit 9, and a CCD sensor 10. The reproduction optical system 5 is provided below the reading optical system 4, and has a laser driver unit 11 and mirrors 12 and 12.

In the reading optical system 4 described above, the original document (not shown) placed on the original document table 2 by the exposure lamp 7.
When the light is irradiated on, the reflected light from the document is guided by, for example, mirrors 8 ... 8 as shown by an arrow, passes through the lens unit 9, and reaches the CCD sensor 10. The reflected light is converted by the CCD sensor 10 into an electric signal for each of red, green and blue color components. The reading optical system 4 is movably provided so that the exposure lamp 7, the mirrors 8 ... 8 and the lens unit 9 irradiate the image of the document with light.
It is provided so as to guide the reflected light from the document to the fixed CCD sensor 10.

On the other hand, in the reproduction optical system 5, each electric signal output from the CCD sensor 10 is converted into a laser beam by the laser driver unit 11, and the laser beam is guided to the copying process section 6 by the mirrors 12 and 12. .

Copying process section 6 as image forming means
Is the photoconductor drum 13, the charger 14, the developing tank 1.
5 to 18, a transfer device 19, a cleaning device 20, a charge eliminating lamp 21, a conveyor belt 22, a fixing unit 23 and the like. The photoconductor drum 13 is formed in a drum shape,
It is rotationally driven in the direction of arrow A. Around the photoconductor drum 13, a charging charger 14, developing tanks 15 to 18, a transfer device 19, a cleaning device 20, and a discharge lamp 21 are arranged.

The charging charger 14 is a charging device that charges the surface of the photosensitive drum 13 prior to exposure. The developing tank 15 contains black toner and is a device for developing the electrostatic latent image formed on the photosensitive drum 13 into black and white. The developing tanks 16 to 18 are yellow,
It contains magenta and cyan toners individually.
This is a device for developing the electrostatic latent image formed on the photosensitive drum 13 into each of those colors.

The transfer device 19 has a transfer belt 24, rollers 26 ... 26, first transfer rollers 27 and 27, and a second transfer roller 28.

The transfer belt 24 is pressed against the photoconductor drum 13 and the three rollers 26, ...
It is supported so as to be rotatable in any direction, and the color toner image formed on the photosensitive drum 13 is transferred. The first transfer roller 27 is arranged on the back surface side of the transfer belt 24, and transfers the toner image on the photosensitive drum 13 to the transfer belt 24. The second transfer roller 28 is disposed on the front surface side of the transfer belt 24,
The toner image on the transfer belt 24 is transferred onto a transfer paper.

The cleaning device 20 includes the photosensitive drum 1
3 is a device for removing the toner remaining on the surface of the toner. The discharge lamp 21 is a device that removes the potential remaining near the surface of the photosensitive drum 13. The conveyor belt 22 is a device that conveys the transfer paper on which the toner image is transferred, and is provided near the second transfer roller 28. The fixing device 23 is a device that thermally fixes the toner on the transfer paper conveyed by the conveyor belt 22.

Copy process unit 6 configured as described above.
In, the image formation is executed by the following procedure.
First, when the photoconductor drum 13 is uniformly charged by the charging charger 14, the reading optical system 4 performs the first image reading. At this time, the image signal (R, G, B) read by the CCD sensor 10 becomes yellow data in an image processing unit (not shown) and is given to the laser driver unit 11.

From the laser driver unit 11,
Laser light corresponding to yellow data is emitted, and the photoconductor drum 13 is exposed by this laser light. As a result, an electrostatic latent image corresponding to the yellow image is formed on the photoconductor drum 13. Next, the developing tank 16 is formed on the electrostatic latent image.
The yellow toner is supplied from the above, and a yellow toner image of the same color is formed.

Then, the yellow toner image is circularly moved in the direction of arrow A and is transferred onto the transfer belt 24 by the first transfer rollers 27. At this time, a part of the toner remains on the surface of the photosensitive drum 13 without being transferred, but this toner is removed by the cleaning device 20. Then, the electric charge remaining on the surface of the photoconductor drum 13 is removed by the static elimination lamp 21.

When the above steps are completed, the charging charger 14 uniformly charges the photosensitive drum 13. The image signal obtained by the CCD sensor 10 is read by the second image reading, and this image signal is given to the laser driver unit 11 as magenta data via the image processing section.

Thereafter, the photosensitive drum 1 is processed in the same manner as described above.
An electrostatic latent image corresponding to the magenta image is formed at 3. By supplying magenta toner from the developing tank 17 to this electrostatic latent image, a magenta toner image is formed. Then, this magenta toner image is transferred to the transfer belt 24 and superposed on the yellow toner image.

After that, the toner is removed and the photoconductor drum 13 is discharged in the same manner as described above, and the photoconductor drum 13 is uniformly charged. Then, the image signal obtained by the third image reading is converted into an electrostatic latent image as cyan data, and the cyan toner in the developing tank 18 forms a cyan toner image.

The cyan toner image is superposed on the transfer belt 24 with the yellow toner image and the magenta toner image to complete the toner image. This toner image is
It is transferred by the second transfer roller 28 onto the transfer paper supplied from the paper feed cassette 29. Further, the transfer sheet is conveyed to the fixing unit 23 by the conveying belt 22, undergoes a fixing process here, and is then ejected outside by the ejecting roller 30.

The above process is a process for forming an image of three colors. However, when forming an image of four colors,
To this, black processing including development with black toner in the developing tank 15 is added. In the case of black-and-white copying, the electrostatic latent image formed on the photoconductor drum 13 is developed with the above black toner, and this toner image is transferred to the transfer paper via the transfer belt 24.

As shown in FIG. 3, the operation unit 3 as the image quality condition input means includes a color adjustment selection key 31, a copy density scale lamp 32, a color adjustment scale lamp 33, a density / color adjustment key 34, an ID key 35 and the like. Is equipped with
Each is used when performing the following operations.

(A) Color adjustment selection key 31: Selection of copy density adjustment or color adjustment (b) Copy density scale lamp 32: Display of density level during copy density adjustment (c) Color adjustment scale lamp 33: Color adjustment Each adjustment level display at the time (d) Density / color adjustment key 34: Each level adjustment at the time of copy density adjustment and color adjustment (e) ID key 35: Registration of ID number (described later) and
Execution of copying by D number Note that the density adjustment is performed simultaneously for each color. On the other hand, in the color adjustment, each color is adjusted independently.

In this copying machine, by operating the above various keys, the copy density and color balance are set to "1" to "7".
It is configured to change in 7 steps. For this reason,
By operating these keys (hereinafter referred to as panel operation), the copy density and color balance can be adjusted according to the original image, and a desired copy image can be obtained.

The operation section 3 is also provided with an automatic mode display lamp 36 and an automatic mode selection key 37.
The automatic mode display lamp 36 is a lamp that lights up in the automatic mode state. The automatic mode selection key 37 is a key for selecting ON / OFF of an automatic mode, which will be described later, and this key enables selection of the automatic mode as needed. In addition, the operation unit 3 is provided with a ten key 38 for setting the number of copies, a copy start button 39, and the like.

[Structure and Operation of Neuro Computer] This copying machine is a neuro computer 41 shown in FIG.
Is used to set the image quality. less than,
The configuration and learning operation of the neuro computer 41 will be described. Note that, here, for convenience, the neurocomputer 41 is described by being replaced with a general three-layer perceptron type neurocomputer.

The neuro computer 41 has an input layer 42.
And an intermediate layer 43 and an output layer 44 have a three-layer structure. The input layer 42 is composed of I units, inputs values measured by the sensor, and outputs the input values as they are to the intermediate layer 43. The intermediate layer 43 is composed of J units similar to the input layer 42, and each unit is coupled to each unit of the input layer 42 via a coupling portion 45 having a unique coupling load. The units of the intermediate layer 43 are not connected to each other.

For example, as shown in FIG. 5A, the output value O2j of the j-th unit of the intermediate layer 43 is the input value of the input layer 4
2, the output value O1i of the i-th unit, the coupling load W1ij from the i-th unit of the input layer 42 to the j-th unit of the intermediate layer 43, and the threshold value W10j of the j-th unit of the intermediate layer 43. Is expressed by the equation (1).

[0065]

[Equation 1]

It should be noted that f is a sigmoid function shown in FIG. 6 represented by the equation (2), which is a non-linear function for determining the input / output characteristics of the neurocomputer 41.

[0067]

[Equation 2]

The output layer 44 is composed of K units similar to the intermediate layer 43, and each unit is connected to each unit of the intermediate layer 43 through a connecting portion 46 having a unique connecting load. For example, as shown in FIG.
The output value O3k of the kth unit of the output layer 44 is the same as the output value O2j of the jth unit of the intermediate layer 43 and the intermediate layer 4
Expression (3) is expressed using the coupling load W2jk from the j-th unit of 3 to the k-th unit of the output layer 44 and the threshold value W20k of the k-th unit of the output layer 44.

[0069]

[Equation 3]

Next, the learning procedure of the neuro computer 41 configured as described above will be described with reference to the flowcharts of FIGS. 7 and 8.

Here, I data is given to I units of the input layer 42 and K data is given to K units of the output layer 44 as teacher data, respectively. As a result, a Max consisting of K data for I data
GN data sets are obtained. However, it is assumed that there is some relationship between the I data and the K data. Further, generally, the connection weights of the connection parts 45 and 46 are determined by learning on a computer from data obtained in advance by experiments. However, in the present embodiment, as will be described later, each coupling load is determined by the amount showing the characteristics of the document and the setting condition of the user.

First, the unit number i of the input layer 42 is set to 0 (S1), and the unit number j of the intermediate layer 43 is set to 1 (S2). The correction amount for the connecting load of the connecting portion 45 is 1
The variable OW1 for saving generations is assigned to each unit number i.
Clear according to j (S3).

The variable OW1 is stored in the TempRAM 76 (FIG. 1) prepared separately from the connection weight memory described later.
9), etc., and is secured in a working memory. Also, OW1
[0] [j] represent each unit of the intermediate layer 43 (hereinafter, as appropriate,
This is an area for storing the correction amount of the threshold value of the intermediate unit). This area is also assigned to the working memory.

Next, 1 is added to j (S4), and it is determined whether or not j exceeds the total number J of intermediate layers 43 (S5). If j is equal to or less than J in S5, the process returns to S3, and if j exceeds J, 1 is added to i (S6). Further, it is determined whether i exceeds the total number I of units in the input layer 42 (S7). If i is less than or equal to I in S7, the process returns to S2, while if i exceeds I, the clearing of OW1 is completed for each unit of the input layer 42 and the intermediate layer 43.

In the subsequent processing, in the same manner as the above processing,
The variable OW2 for storing the correction amount of the coupling load of the coupling unit 46 for one generation is cleared for each unit of the intermediate layer 43 and the output layer 44 (S8 to S1).
4).

The variable OW2 is also secured in the working memory as is the case with the variable OW1. Also, OW2
[0] [k] represents each unit of the output layer 44 (hereinafter, as appropriate,
This is an area for storing the threshold correction amount of the output unit). This area is also assigned to the working memory.

After that, the error storage variable r for judging the end of learning is initialized to 0 (S15), and GrpNo indicating the set number of the measurement data is initialized to 1 (S16).
When this is completed, i is set to 1 (S17), and the input layer 42
To each unit (hereinafter appropriately referred to as an input unit) of
The data indicated by rpNo is input (S18).

Further, 1 is added to i (S19), and it is determined whether i exceeds I (S20). S20
If i is less than or equal to I, the process returns to S18, while i
When is larger than I, it means that the data input is completed for all the input units.

When the data input is completed, the output value of each intermediate unit is calculated according to the equation (1). In this process, first, j is set to 1 (S21), and the output value O2j of the first intermediate unit is set to -W1 [0] [j] (threshold of the j-th intermediate unit) (S22). Here, i is set to 1 again (S23), and O1 [i] and W1 are set.
A value obtained by adding the output value obtained as described above to the product of [i] and [j] is set as the output value of each new intermediate unit (S2).
4).

Further, 1 is added to i (S25), and it is determined whether i exceeds I (S26). S26
If i is less than or equal to I, the process returns to S24, while i
Is larger than I, it means that all the output values of the first intermediate unit based on the output value from each input unit have been obtained.

Then, calculation is performed by the sigmoid function of the equation (2) using O2j obtained in S24 as a parameter (S27). Then, 1 is added to j (S28), and it is determined whether j exceeds J (S29). S29
If j is less than or equal to J, the process returns to S22, while j
Is greater than J, it means that the output values of all the intermediate units up to the Jth have been obtained.

When the output value of each intermediate unit is obtained,
Outputs of all output units up to the K-th are obtained according to the equation (3) in the same procedure as described above (S30 to S38). Note that W2 [0] [k] in the processing of S31 is the threshold value of the kth output unit.

When the above processing is completed, output data corresponding to the input data currently being calculated is input as teacher data (Dt) for each output unit (S39 to S4).
2). Then, k is set to 1 (S43), the squared error between the value obtained in the output layer 44 and the above-mentioned teacher data is calculated for each output unit, and the result is added to r to be a new r (( S44). Then, the error e3 [k] for correcting the coupling load of the coupling section 46 is obtained according to the equation (4) for each output unit (S45 to 47).

[0084]

[Equation 4]

However, in the above equation, k = 1 to K.

Similarly, an error e2 [j] for correcting the coupling load of the coupling portion 45 for each intermediate unit is obtained.
Is calculated according to the equation (5) (S48 to 55).

[0087]

[Equation 5]

However, in the above equation, j = 1 to J. Further, W2 [j] [k] indicates the coupling load between the jth intermediate unit and the kth output unit.

In the subsequent processing, the coupling load W2 [j] [k]
The correction amount dW2 [j] [k] is calculated according to the equation (6) (S58).

[0090]

[Equation 6]

However, in the above equation, the coefficient ε is a minute value and is usually set to a magnitude of about 0.1.

When dW2 [j] [k] is obtained, the coupling weight W2 [j] [k] is corrected using this (S59).

[0093]

[Equation 7]

However, in the above equation, the coefficient α is a minute value, and is set to the same magnitude as the coefficient ε.

Further, dW2 [j] [k] at the present time is set to OW2 [j] [k] for use in the next calculation (S60). All of the coupling loads 56 are corrected by the above procedure (S56 to S64), and the coupling loads W1 [i] [j] of the coupling portion 45 are similarly corrected (S65 to 73).
The calculation for this correction is executed according to the equations (8) and (9).

[0096]

[Equation 8]

[0097]

[Equation 9]

Then, the processing of S17 to S73 is performed on the MaxGN tutor data (S74, S75).
After that, if r is less than the value R for judging the end of learning, it is considered that the neurocomputer 41 has learned the relationship between the input data group and the output data group for the teacher data (S76), and the processing ends. If r is R or more, the processing from S15 is repeated.

In the neuro computer 41 which has completed the learning as described above, a processing procedure for outputting an expected value for given data will be described with reference to the flowchart of FIG.

First, input data is set in I input units (S101 to S104), and the output value of each intermediate unit is calculated according to the equation (1) (S105 to S11).
3). Next, the output value of each output unit is calculated according to the equation (3) (S114 to S122). The above calculation is the same as the processing of S18 to S38 in the learning procedure described above.

After the learning is completed, the neuro computer 41 obtains the relationship between the input / output data for the teacher data as the connection weight of the connection parts 45 and 46, and is unique to the teacher data for an appropriate input value. Output the value expected from the input / output relationship. In addition, the neuro computer 41
Operates almost correctly for input values that have not been learned, and outputs an almost accurate expected value for any input value if the input value range is appropriate.

This copying machine is equipped with a processing system for image quality setting, which is equipped with the neuro computer 41 described above. In this embodiment, the following seven examples will be described.

[First Processing System] As shown in FIG.
In the above processing system, the image signals (R, G, B) output by the three CCD sensors 10 ...
.. are converted into digital image data (8 bits) and are given to the histogram counters 52.

The histogram counter 52 carries out a histogram calculation process for distributing the density of the above image data in the range of 0.0 to 1.0 at intervals of 0.2 as shown in FIG. The histogram counter 52 is configured to generate a total of 15 pieces of histogram data, five for each color of red, green and blue by performing the above-mentioned histogram calculation processing.

The histogram counter 52 is, for example, as shown in FIG.
1, the comparators 101 to 104 and the ROM 10
5 and counters 106 to 110.

Each of the comparators 101 to 104 has an input I.
Image data of 8 bits is input to ₁ as a comparison value, while values of 0.2, 0.4, 0.6 and 0.8 are input to the input I ₂ as reference values. Comparators 101-104
Compares the comparison value with the reference value, and outputs "1" when the comparison value is smaller than the reference value, and outputs "0" otherwise.

The ROM 105 has address inputs A _{0 to} A ₃
The outputs of the comparators 101 to 104 are connected to 0.
0-0.2, 0.2-0.4, 0.4-0.6, 0.6
The addresses shown in Table 1 are assigned to each of the five distribution patterns of 0.8 to 0.8 and 0.8 to 1.0. ROM105
Outputs "1" when an address corresponding to the value given to the address inputs A _{0 to} A ₃ is designated. For example, when the comparison value is 0.7, only the value given to the address input A ₃ becomes “1”, and the distribution pattern corresponding to the address “0001” designated by this is 0.6 to 0.8. Value becomes "1".

[0108]

[Table 1]

The counters 106 to 110 are respectively RO
Five outputs (of a distribution pattern) of M105 are individually input to the clock input CK. That is, the counters 106 to 110 count “1” that is input and output the count value as histogram data of each distribution pattern.

The histogram counter 52 is configured to generate the histogram data and take the characteristic of the image data for each pixel by the above configuration. Since it can be considered that changes in the density of the original image do not change rapidly between neighboring pixels, the histogram data does not have to be generated for all pixels. For example, the histogram data may be extracted by thinning out every 8 pixels to roughen the characteristics of the pixel data. In this way, the number of processes can be reduced and the processing speed can be increased.

In this copying machine, the image information (histogram data) of the original image is read by the image information reading section 57 composed of the CCD sensors 10 ... 10, the A / D converters 51 ... 51 and the histogram counters 52 ... 52.
Is designed to read. That is, the image information reading unit 57 has a function as an image information reading unit. Further, the above-mentioned histogram data is given to the learning processing unit 58 as information regarding the color of the document image that is correlated with the image quality condition set by the user.

Learning processing unit 58 as means for determining image quality conditions
Is a plurality of neuro computers 41 ... And a switch 6
1 and 62, and each neuro computer 41
The above histogram data is input to ... The neuro computer 41 may be configured by a dedicated LSI or the like, or may be configured by programming a commonly used microprocessor or the like, for example.

The neuro computer 41 performs the above-described learning based on the input histogram data and the setting of the image quality condition by the operation unit 3, and at the time of copying, based on the histogram data obtained at that time, the learning content is changed. Therefore, the image quality data given to the internal parameter section 59 is output. The image quality data corresponds to copy density and color balance.

Further, each neuro computer 41 ...
Data is exchanged with a memory card 60 as an external storage medium. In the memory card 60, the coupling unit 45 obtained by the learning by the neurocomputer 41
-46 coupling loads are stored.

The output side of the neuro computer 41 is selected via the switch 61. The switch 61 is switched by the ID code number output from the operation unit 3, and is further connected to the switch 62.

The switch 62 includes the switch 61 and the operating section 3
Alternatively, the connection with the internal parameter section 59 is switched. This switch 62 is for the neurocomputer 4
The switch 61 and the operation unit 3 are connected to each other during learning of 1, while the switch 61 and the internal parameter unit 59 are connected to each other during copying.

As described above, the operation section 3 as the image quality condition input means is constructed so that the image quality conditions of the copy density and the color balance can be set in each of seven levels. Specifically, depending on the operation unit 3 to 7 steps,
Seven types of image quality data up to 0.2, ..., 0.7 are output as image quality conditions and given to the internal parameter section 59. In addition, the operation unit 3 outputs the ID personal identification number. The ID code is, for example,
The number is arbitrarily set by the user and is registered in advance.

In the copying machine of the present embodiment, the mode in which the copy image quality for each user is set using the ID code is called the ID copy mode, and the mode in which the image quality is automatically set using the neurocomputer 41 is called the automatic mode. I'm calling. In the initial state of the automatic mode, both copy density and color balance are set to "4", which is an intermediate value. Further, in the automatic mode, the coupling load of the neuro computer 41 is set so that the image quality data of 0.4 is output in the initial state. The ID key 35 is operated to shift from the normal mode in which normal copying is performed to the ID copy mode. And if it becomes ID copy mode,
By inputting the ID code with the ten keys 38, the neuro computer 41 corresponding to the ID code is selected.

Internal parameter section 59 as a conversion means
Includes a memory, and when the image quality data is input from the operation unit 3 or the neuro computer 41, the internal parameters stored in the memory are output according to the image quality data. Specifically, the internal parameter unit 59 is input by the user from the operation unit 3 at the time of learning of the neuro computer 41 by 0.1, 0.2.
A value suitable for a basic control pattern of process control is output as an internal parameter according to the image quality data of 0.7.

That is, the image quality data is a value set stepwise at equal intervals so that the user can easily recognize it on the display of the operation unit 3, but this value is not suitable for complicated process control. Therefore, it is necessary to convert the image quality data into internal parameters suitable for process control.
Therefore, the memory serves as a conversion table for converting the image quality data into internal parameters.

The operation section 3 and the internal parameter section 59 are connected via a switch 64. This switch 64
ON / OFF of the operation unit 3 and the internal parameter unit 5
9 is connected or disconnected. Switch 64
Is ON during learning of the neuro computer 41 or during normal copying without using the neuro computer 41, while O is turned on during copying using the neuro computer 41.
It is supposed to FF.

The process control section 63 as the control means is
Based on the above internal parameters, add other correction factors etc. as calculation parameters to calculate an appropriate process control value,
Each part of the copying process unit 6 is controlled by the process control value. As the process control value, for example, the exposure amount of the exposure lamp 7, the charging output of the charging charger 14, the bias voltage of the developing tanks 15 to 18, the transfer device 1
The transfer output of No. 9 and the amount of static elimination light of the static elimination lamp 21 are used.

The operation of the first processing system configured as described above during learning will be described.

First, when the ID personal identification number is registered by the input from the operation unit 3, one neurocomputer 41 is assigned to the ID personal identification number, and the switch 61 is switched to the neurocomputer 41. During learning, switch 61
Is connected to the operation unit 3, and the switch 64 is turned on.
I am N.

On the other hand, the 15 pieces of histogram data from the histogram counters 52 ... 52 are input to the input side of the neuro computer 41. When the image quality condition set by the operation unit 3 is output as image quality data, this image quality data is input to the output side of the neuro computer 41 via the switch 62. Then, the neuro computer 41 uses the image quality data as teacher data,
The relationship between the histogram data and the image quality data is learned by the procedure shown in FIGS. 7 and 8.

At this time, the image quality data is also given to the internal parameter section 59 through the switch 64. Therefore, the internal parameter section 59 outputs an internal parameter according to the image quality data, and the process control section 63 gives a process control value to the copy process section 6.

The copying operation after learning by the first processing system is
This will be described with reference to the flowchart of FIG.

When the copy start button 39 is turned on, first, the image information reading section 57 reads the image information of the original image (S201) and determines whether or not the ID copy mode is set (S202). In S202, in the ID copy mode, the switch 61 is switched according to the ID code number, and the learning computer 58 selects the neurocomputer 41 corresponding to the ID code number (S203).

Then, it is judged whether or not the automatic mode is selected (S
204). In the automatic mode in S204, based on the image information read by the image information reading unit 57, according to the learning content of the neuro computer 41, as shown in FIG.
The image quality data is calculated by the above-described procedure shown in FIG. 5, and the internal parameters are automatically set based on this image quality data (S20).
5).

At this time, the neuro computer 41 and the internal parameter section 59 are connected via the switch 62, and the switch 64 is off. Therefore,
The internal parameter section 59 includes a neuro computer 41.
Only the image quality data from (1) to (4) are given as image quality conditions through the switch 62. Then, the internal parameter section 59
Outputs internal parameters according to image quality data. Then, when this internal parameter is given to the process control unit 63, the process control unit 63 outputs a process control value according to the internal parameter, and the copy process unit 6 executes copying (S206).

If the automatic mode is not selected in S204, the neurocomputer 41 determines R, obtained based on the image information.
The relationship between the histogram data for each of G and B and the image quality data input from the operation unit 3 is learned (S207). At this time, since the switch 64 is ON, the image quality data from the operation unit 3 is input to the internal parameter unit 59.
As a result, the internal parameter unit 59 determines the internal parameters according to the image quality condition set by the operation unit 3 (S208).

At this time, the image quality conditions set in the operation unit 3 are the standard image quality conditions automatically set when the copying machine is powered on and in the initial state, or the image quality according to the user's operation input. It is a condition. At this time, if the automatic mode is set, the automatic mode is canceled in order to determine the internal parameters by the setting of the operation unit 3.

Then, the copying process unit 6 executes copying based on the internal parameters (S206). In addition,
Although the learning in S207 is not related to the copying executed in S206, the learning is performed for the purpose of repeating the learning to set the image quality according to the user's preference.

If the ID copy mode is not set in S202,
The image quality condition is not set using the neuro computer 41, and the process proceeds to S206 to execute copying.
At this time, the switch 64 is ON, the internal parameter is determined by the image quality data output from the operation unit 3 by the operation of the user, and the copy process unit 6 is controlled based on the internal parameter.

As described above, in the copying machine equipped with the first processing system, when the image quality condition for the copied image is designated, the neuro computer 41 determines the image quality corresponding to the original image and the copied image quality set by the user for the original image. It is designed to learn the correspondence with the data. Thus, at the time of copying, the neuro computer 41 sets an appropriate image quality for the image information input according to the learning content. Therefore, the user does not have to enter the image quality condition for each copy, and the operability is improved.

Also, the number of data handled by this processing system is
By using the neuro computer 41, there are only the number of couplings of the coupling units 45 and 46 × the number of ID personal identification numbers. Therefore, it is not necessary to provide a huge capacity of memory as in the case of storing image information and image quality conditions in a one-to-one correspondence.

In the setting of the copy image quality using the neuro computer 41, the exposure amount can be brought closer to a more accurate value than the setting of the copy image quality by the above-mentioned one-to-one correspondence. As will be described below, this is apparent by comparing the calculation result of the exposure amount with the setting by the neuro computer 41 and the setting of 1: 1.

As shown in FIG. 13, it is assumed that the user's preference of the exposure amount with respect to the average value of the densities of a certain color of the document is as indicated by the solid line. Then, for example, after the patterns in which the average values of the document densities are P, Q, and R are learned or stored by the user's designation, the document image with the average document density of X is copied. In this case, one to one
In the setting of, since the above three patterns are only stored, the exposure amount between the patterns corresponding to Q and R is a value on the broken line connecting these patterns. Therefore, with the one-to-one setting, the exposure amount with respect to the average value X becomes x ′, and the exposure amount cannot be smoothly approximated along the solid line.

When the number of parameters is small as described above, x'is obtained by performing interpolation using a spline function or the like.
Can be approximated to a more accurate value. However, in actual copying, there are many parameters such as the exposure amount, the charging voltage of the photosensitive drum 13, the bias voltage of the developing tanks 15 to 18, the transfer voltage, and the 15 histogram data from the histogram counters 52 ... 52. Therefore, the 1: 1 setting makes it impossible to accurately approximate the exposure amount to the user's preference. Further, in the one-to-one setting, there is a disadvantage that the data cannot be retrieved instantaneously unless the data structure for storage is devised.

On the other hand, in the setting by the neuro computer 41, as shown in the figure, the exposure amount can be approximated to the user's preference, and the expected exposure amount "x" with respect to the average value X can be accurately calculated. be able to. Further, if the neuro computer 41 is used, the calculation is performed instantaneously, so that it does not take time to retrieve the data.

Further, in this processing system, since the learning processing section 58 is provided to perform copying in the ID copy mode, the setting of the copy image quality can be managed for each user by the ID personal identification number. Therefore, the copy image quality settings made by a plurality of users do not coexist in one neurocomputer, and the preference of the copy image quality of each user can be reflected in the copy.

Further, in the present processing system, the learning contents of the learning processing unit 58 are stored in the memory card 60, so that the information stored in the memory card 60 can be stored in a copying machine having another similar function. By providing the image quality condition, it becomes possible to set the image quality condition without additional learning even in the copying machine. Further, if the memory card 60 is provided for each user, each user can perform copying in the ID copy mode by another copying machine.

<Modification> Here, a modification of the present processing system will be described.

In this modified example, as shown in FIG. 14, each neurocomputer 41 ... Communicates with a memory card 65 capable of recording an ID code. The memory card 65 serving as an external storage medium includes a coupling unit 45, which is acquired by the neurocomputer 41 through learning.
The coupling load of 46 is stored and the ID code number is set.

The connection between the memory card 65 and each neurocomputer 41 ... Is switched by a switch 66. The switch 66 is for the memory card 6
The connection can be switched by the ID personal identification number which is different for each 5. Similarly, the switch 61 also has an ID
The number of the neuro computers 41 ... Is switched according to the number.

In the second processing system configured as described above, when the memory card 65 is replaced, the switch 61.
66 is switched. As a result, the neuro computer 41 corresponding to the ID personal identification number is selected.

Therefore, according to the present processing system, by connecting the memory card 65, the neurocomputer 41 is automatically selected, and it is possible to save the trouble of inputting the ID personal identification number from the operation unit 3. Also, the memory card 6
Switch 66 for connecting the computer 5 and the neuro computer 41
Since it is designed to be carried out via a memory card, it is only necessary to provide one connector for connecting a memory card.

[Second Processing System] The same reference numerals are used for the constituent elements in this processing system and the constituent elements in other processing systems described later that have the same functions as those in the first processing system. And its description is omitted.

As shown in FIG. 15, this processing system is different from the first processing system in that the output of the learning processing unit 58 becomes an internal parameter. Therefore, the learning processing unit 58 in this processing system functions as a parameter determining unit.

Specifically, the operation unit 3 and the internal parameter unit 59 are directly connected, and the connection between the switch 61 and the output side of the internal parameter unit 59 or the input side of the process control unit 63 is switched by the switch 67. ing.
The switch 67 connects the switch 61 and the internal parameter section 59 during learning of the neuro computer 41, and connects the switch 61 and the process control section 63 during copying.

Further, the internal parameter section 59 and the process control section 63 are connected via a switch 68.
The switch 68 is turned on during learning of the neuro computer 41 or during normal copying without using the neuro computer 41, and is turned off during copying using the neuro computer 41.

By the way, the neuro computer 41 ...
Communicate with the memory card 60 via the switch 66. The switch 66 is switched by the ID personal identification number output from the operation unit 3.

The operation of the second processing system configured as described above during learning will be described.

First, when the ID personal identification number is registered by the input from the operation unit 3, one neurocomputer 41 is assigned to the ID personal identification number, and the switch 61 is switched to the neurocomputer 41. Further, at the time of learning, the switch 67 is used to switch the switch 61.
And the internal parameter section 59 are connected, and the switch 68 is turned on. At this time, the internal parameter section 59 outputs an internal parameter based on the image quality data output from the operation section 3 by the setting by the user or automatically.

On the other hand, the 15 pieces of histogram data from the histogram counters 52 ... 52 are input to the input side of the neuro computer 41, and the above internal parameters are input to the output side of the neuro computer 41 via the switch 67. Then, the neuro computer 41
Learns the relationship between the histogram data and the image quality data in the procedure shown in FIGS. 7 and 8 using the image quality data as teacher data.

At this time, the above internal parameters are also given to the process control unit 63 through the switch 68. Therefore, the process control unit 63 outputs the process control value according to the internal parameter, and the copy process unit 6
Given to.

The copy operation after learning by the second processing system is
This will be described with reference to the flowchart of FIG. The copying operation in this case is basically performed in the same procedure as the copying operation by the first processing system.

When the copy start button 39 is turned on, first, the image information reading unit 57 reads the image information of the original image (S301), and it is determined whether or not the ID copy mode is set (S302). If the ID copy mode is selected in S302, the neurocomputer 41 corresponding to the ID code number is selected (S303), and it is determined whether the automatic mode is selected (S304). In S304, in the automatic mode, internal parameters are calculated based on the read image information according to the learning content of the neuro computer 41, and the process control value is automatically set by this internal parameter (S305).

If the automatic mode is not selected in S304, the neuro computer 41 determines the histogram data for each of R, G, and B obtained based on the image information and the input from the operation unit 3 in the internal parameter unit 59. The relationship with internal parameters is learned (S307). At this time, since the switch 68 is ON, the internal parameter from the parameter 59 is input to the process control unit 63. As a result, the process control unit 63 determines the process control value according to the image quality condition set by the operation unit 3 (S
308).

If the automatic mode is set in S308, the automatic mode is canceled in order to determine the process control value by the setting of the operation unit 3. Then, when the above process control value is given to the copy process unit 6, the copy process unit 6 executes copying based on this process control value (S306).

Then, the copying process unit 6 executes copying based on the process control value (S306). In addition,
Although the learning in S307 is not related to the copy executed in S306, the learning is performed for the purpose of repeating the learning to set the image quality according to the user's preference.

If the ID copy mode is not set in S302, the image quality condition is not set using the neuro computer 41, and the process proceeds to S308 to execute copying. At this time, the switch 68 is ON, the process control value is determined by the internal parameter output from the internal parameter 59, and the copying process unit 6 is controlled based on the process control value.

As described above, in the copying machine having the second processing system, when the image quality condition for the copied image is designated, the neurocomputer 41 responds to the image information of the original image and the copied image quality set by the user. It is designed to learn the correspondence with internal parameters. Thus, at the time of copying, the neuro computer 41 sets an appropriate image quality for the input image information according to the learning content. Therefore, similar to the copying machine having the first processing system, the operability can be improved, and it is not necessary to have a huge memory for storing the above correspondence.

Since the connection between the memory card 60 and the neuro computer 41 is made via the switch 66, one connector for connecting the memory card is used.
You only need to provide one.

Also in this processing system, the memory card 60 may be replaced with the memory card 65 as in the first processing system. Also in this case, the ID by the memory card 65
The neuro computer 41 can be used according to the personal identification number.

[Third Processing System] As shown in FIG. 17, this processing system is different from the first and second processing systems in that a plurality of neuro / buffer sections 69 and switches 61, 62, 71 are provided.
The learning processing unit 70 having The learning processing unit 70 as the image quality condition determining unit is connected to the histogram counters 52 ... 52 via a switch 71. The switch 71 is adapted to be switched by the ID personal identification number from the operation unit 3.

The neuro / buffer unit 69 includes a neuro computer 41 and a buffer RAM as shown in FIG.
72 and. In addition, the neuro computer 41
A switch 73 is provided between the buffer RAM 72 and the switch 71. On the other hand, a switch 74 is provided between the neuro computer 41, the buffer RAM 72 and the switch 61.

The switch 73 has three contacts 73a-73.
c, and the connection between the contacts 73a to 73c is switched according to the following three cases.

(A-1) Preparation stage of learning of the neuro computer 41 The contacts 73a and 73b are connected. As a result, the histogram data input via the switch 71 is guided to the buffer RAM 72.

(A-2) During learning of the neuro computer 41 The contacts 73a and 73b are separated, and the contacts 73b and 73c are connected. As a result, the buffer RAM 7
The histogram data accumulated in No. 2 is guided to the neuro computer 41.

(A-3) During execution of copying in automatic mode The contacts 73a and 73c are connected. This allows
The histogram data input via the switch 71 is directly input to the neuro computer 41 without passing through the buffer RAM 72.

The switch 74 has three contacts 74a-74.
c, and the connection between the contacts 74a to 74c is switched according to the following three cases.

(B-1) Preparation stage of learning of the neuro computer 41 The contacts 74a and 74b are connected. As a result, the image quality data input via the switch 61 is transferred to the buffer RA.
Guided by M72.

(B-2) At the time of learning of the neuro computer 41, the contacts 74a and 74b are separated and the contacts 74b and 74c are connected. As a result, the buffer RAM 7
The image quality data accumulated in 2 is stored in the neuro computer 41.
Be led to.

(B-3) During execution of copying in automatic mode The contacts 74a and 74c are connected. This allows
The image quality data output from the neuro computer 41 is output via the switch 61.

The neuro computer 41, as shown in FIG. 19, has a ROM 75, a TempRAM 76, and a CP.
It is equipped with U77. The ROM 75 is a memory that stores a program or the like for the CPU 77 to execute a neuro operation. The TempRAM 76 is a memory that stores data about the current document. CPU77 is R
According to the program stored in the OM 75, R stored in the buffer RAM 72 as described later,
A neuro calculation is performed based on each of the G and B histogram data. Further, although not shown, the neuro computer 41 is provided with a connection weight memory, and the connection weight obtained as a result of learning is stored in this connection weight as learning content.

Incidentally, in FIG. 19, for convenience, the switches 73 and 74 shown in FIG. 18 are omitted.

On the other hand, the buffer RA as a data storage unit
As shown in FIG. 20, M72 is a histogram data of five R, G, and B histograms and R (C), G corresponding thereto.
The density values (image quality data) of (M) and B (Y) are, for example, 1
00 sets can be stored.

The learning operation of the third processing system configured as described above will be described.

First, when the ID code number is registered by input from the operation section 3, one neuro / buffer section 69 is assigned to the ID code number, and the switch 61 is switched to the neuro / buffer section 69. To be During learning, switch 61
Is connected to the operation unit 3, and the switch 64 is turned on.
I am N.

On the other hand, 15 pieces of histogram data from the histogram counters 52 ... 52 are input to the input side of the neuro / buffer section 69 through the switch 71.
In addition, the image quality condition set by the user from the operation unit 3 is set as image quality data through the switch 61.
It is input to the output side of the buffer unit 69. Then, in the neuro / buffer unit 69, both of the above data are transferred to the switch 7
It is written in the buffer RAM 72 through 3.74 and stored as learning data. At this time, the buffer RAM 7
In the case of 2, when 100 sets of learning data can be stored, for example, when the number of data of learning data exceeds 100, the oldest data is deleted in order.

Then, the accumulated n sets of learning data are given to the neuro computer 41 through the switches 73 and 74. The neuro computer 41 learns the relationship between the histogram data and the image quality data based on the learning data in the procedure shown in FIGS. 7 and 8.

Also, the above image quality data is stored in the switch 64.
Through the internal parameter section 59. Therefore, the internal parameter section 59 outputs an internal parameter corresponding to the image quality data, and the process control section 63 further outputs the internal parameter.
A process control value is given to the copying process unit 6 from the.

The copy operation after learning by the third processing system is
This will be described with reference to the flowchart of FIG.

When the copy start button 39 is turned on, first, the image information reading unit 57 reads the image information of the original image (S401) and determines whether or not the ID copy mode is set (S402). If the ID copy mode is selected in S402, the switch 61 is switched according to the ID code number, and the learning / processing section 70 selects the neuro / buffer section 69 associated with the ID code number (S40).
3).

Then, it is judged whether or not the automatic mode is selected (S
404). If the automatic mode is set in S404, the internal parameters are set based on the image information read by the image information reading unit 57 according to the learning content of the neurocomputer 41 according to the procedure shown in FIG. 9 (S405). ).

At this time, a neuro /
The buffer unit 69 and the internal parameter unit 59 are connected, and the switch 64 is off. In the neuro / buffer unit 69, the switch 73 connects the contact points 73a and 73c, and the switch 74 connects the contact point 7a.
4a and the contact 74c are connected (S405). As a result, the neuro computer 41 is connected to the histogram counters 52 ... 52 and the internal parameter section 59.

When the histogram data from the histogram counters 52 ... 52 is given in this state, the neuro computer 41 determines the image quality data suitable for the histogram data according to the learning content. Then, the internal parameter section 59 outputs internal parameters according to the image quality data output from the neuro computer 41. In this way, the neurocomputer 41 automatically sets the internal parameters (S406).

This internal parameter is the process control unit 63.
Is given to the process control unit 63, the process control unit 63 outputs a process control value according to the internal parameter, and the copy process unit 6 executes copying (S407).

If it is determined in S404 that the buffer RAM 72 can store 100 sets of learning data when not in the automatic mode, it is determined whether the number of data is 100 sets or less (S408). If the number of data is 100 or less in S408, the switch 73 connects the contact points 73a and 73b, and the switch 74 connects the contact points 74a and 74b (S409). This allows R,
The histogram data for each of G and B and the image quality data from the operation unit 3 are written in the buffer RAM 72 through the switches 73 and 74, respectively, and the learning data is registered (S410).

Thereafter, the switch 73 connects the contact points 73b and 73c, and the switch 74 changes the contact point 74.
b and the contact 74c are connected (S411). As a result, the neuro computer 41 becomes
The learning data registered in is given, and the relationship between the histogram data and the image quality data is learned (S412).

When learning is completed, the user operates the operation unit 3
The internal parameters are determined by the internal parameter unit 59 based on the image quality data from (S413). Then, copying is executed based on the internal parameters (S407).

If the number of data exceeds 100 in S408, the switch 73 is set to the contact 7 as in S409.
3a and the contact 73b are connected, and a switch 74
Connects the contact point 74a and the contact point 74b (S414).
In this case, the oldest learning data is rewritten and registered in the buffer RAM 72 as new learning data (S41).
5).

Then, similarly to S411, the switch 73
Connects the contact points 73b and 73c, and the switch 74 connects the contact points 74b and 74c (S4).
16). Then, the neuro computer 41 learns the relationship between the histogram data and the image quality data by the learning data registered in the buffer RAM 72 (S41).
7).

When the neuro computer 41 finishes the learning, the process proceeds to S413. On the other hand, in S402,
If the ID copy mode is not set, the process similarly shifts to S413.

As described above, since the copying machine equipped with the third processing system is equipped with the buffer RAM 72, it is possible to learn many data. As a result, learning that is biased only to specific image information is not performed, and more accurate image quality conditions can be set during copying.

In the control system which does not include the buffer RAM 72, the learning is repeated until the r becomes smaller than R in the processing of S76 in the learning procedure shown in FIGS. Is determined. Therefore, it is possible to obtain an accurate joint weight for the newest data, but since the learning is not repeated for the already input data as described above, the joint weight may be incorrect. was there.

However, as described above, the buffer RA
If M72 is provided, learning is repeated until r becomes smaller than R for all data in the buffer RAM 72. Therefore, an accurate coupling weight can be obtained for all data.

Although the present processing system is designed to learn the relationship between the histogram data and the image quality data, like the second processing system, the relationship between the histogram data and the internal parameters is learned. It may be one.
In this case, the learning processing unit 70 functions as a parameter determining unit.

Further, the present processing system may be constructed so that the number of sheets used and the like can be managed for each ID personal identification number, or copying cannot be executed unless the ID personal identification number is registered.

<Modification 1> In the first modification of the present processing system, learning is not performed until a certain amount of learning data is accumulated in the buffer RAM 72. Specifically, assuming that the capacity of the buffer RAM 72 is, for example, 100 sets, the neuro computer 41 performs learning after 100 sets of data are accumulated.
Further, when the learning data exceeds 100 sets, the old learning data is deleted in order and new learning data is registered.

In the copying operation after learning according to this modification,
In the flowchart of FIG. 21, S408 and S409
The process shown in FIG. 22 is added between and.

In this processing, the number of data is 10 in S408.
After determining that the number of pairs is 0 or less, the data is exactly 100.
It is determined whether or not it is a group (S418). In S418,
If there are 100 sets of data, the process proceeds to S409. On the other hand, when the number of data is less than 100 in S418, the switch 73 connects the contact points 73a and 73b, and the switch 74 connects the contact points 74a and 74b as in S409 (S419). Then, the histogram data and the image quality data obtained at that time are registered in the buffer RAM 72 as learning data (S
420), the process proceeds to S413.

By the above processing, the learning data is 10
If the number of sets is less than 0, only the learning data is registered, and 100 sets of the learning data are accumulated and the learning is performed in a sufficiently secured state. As a result, it is possible to prevent the learning from being performed while the learning data is insufficient.

If learning is performed with a small amount of specific learning data, the amount of correction of the coupling weight may be large when learning is subsequently performed with a large number of learning data. Therefore, more time is required for learning. There are cases. For example, it is assumed that one of the connection weights has a large negative value by learning with a small amount of learning data. However, in order to learn with a large number of learning data, it may be better that the connection weight is positive. In such a case, the coupling weight is relearned from a large negative value to a large positive value more times.

On the other hand, in the processing according to the present modification, since learning is collectively performed after sufficient learning data is secured, learning is not biased to specific learning data, and there is almost no possibility of a long learning time. . Further, since learning is performed with sufficient learning data, the image quality can be set more accurately.

<Modification 2> In the second modification of this processing system, the data writing to the buffer RAM 72 is configured to adopt the condition immediately before the original is exchanged. In the following description, a case will be described in which a method is adopted in which the user performs copying in advance for the same original image with several settings, selects the most preferable image quality condition setting, and performs final copying.

In this case, learning is performed on a plurality of image quality data for the image information of the same document, and the learning does not converge. In the present modification, in order to prevent this, for the image information of the same original document that is continuously input, the neuro computer 41 controls so as to learn only the last input image quality data. In order to realize this, for example, a method of detecting that the cover covering the document placing table 2 is opened or detecting whether or not the document image has changed can be considered.
Here, the latter method will be described with reference to the flowchart of FIG. 23, which shows a control procedure for writing data in the buffer RAM 72.

In this configuration, the TempRAM 76 described above is used.
The memories M ₁ and M ₂ (not shown) are allocated to the memory. The memory M ₁ has an image memory M ₁ (A) for storing histogram data obtained from a document image and a setting memory M ₁ (B) for storing image quality data output from the operation unit 3. . On the other hand, the memory M ₂ has an image memory M ₂ (A) for storing the histogram data obtained from the original image and a setting memory M ₂ (B) for storing the image quality data output from the operation unit 3. ing.

When the power of the copying machine is turned on, first, a flag f indicating whether or not copying is the first time is set to "1" (first time) (S501). When the image quality condition is set by the operation unit 3 (S502), the image quality data output from the operation unit 3 with the setting is stored in the setting memory M ₁ (B) (S
503), copying is executed by the internal parameter determined based on the image quality data (S504). Further, the histogram data at this time is stored in the image memory M ₁ (A) (S505).

Next, it is determined whether the flag f is "1" or "0" (other than the first time) (S506), and if the flag f is "1", it is set to "0" (S507). After that, the histogram data stored in the image memory M ₁ (A) is transferred to the image memory M ₂ (A) (S508). Further, the image quality data stored in the setting memory M ₁ (B) is moved to the setting memory M ₂ (B) (S509), and the process returns to S502.

After that, after repeating the processing from S502 to S503, the second copying is performed in S504, and S506 is performed.
It is determined whether the flag f is "1" or "0" in the above. Here, the flag f is "0" in the previous processing of S507. Therefore, in this case, the data stored in the image memory M ₁ (A) is compared with the data stored in the image memory M ₂ (A), and it is determined whether the two match (S510). ).

If it is determined in S510 that they do not match (the original has been exchanged), the data stored in the image memory M ₂ (A) and the setting memory M ₂ (B) is stored in the buffer R.
It moves to AM72 (S511), and a process transfers to S508. If it is determined in S510 that the two match (the originals are the same), the process proceeds to S509.

In this way, among the image quality conditions successively set for the same original, only the last set image quality condition is registered in the buffer RAM 72. As a result, the data to be truly learned can be given to the neuro computer 41.

<Modification 3> In the third modification of the present processing system, if the data input to the buffer RAM 72 is similar to the data already registered in the buffer RAM 72, the data is stored in the buffer RAM 72. It is configured not to write. That is, T in FIG.
The similarity between the data relating to the current original image stored in the empRAM 76 and the data registered in the buffer RAM 72 of FIG. 20 is determined.

Specifically, the elements of the data of the TempRAM 76 and the data of the buffer RAM 72 are assumed to have different dimensional directivities, and the difference between the values of the elements (dimensions) is regarded as the distance. When the sum a obtained by adding the square of the distance to each element (18 pieces: 15 pieces of histogram data and 3 pieces of image quality data) (hereinafter, simply referred to as sum total a) is smaller than a predetermined value, both data are similar. It is determined that the two data are similar to each other. For example, if the values of 18 elements of both data are Xi and Yi, respectively, the above sum a is
It is expressed by equation (10). Note that i indicates the address where each element is stored.

[0217]

[Equation 10]

Data writing to the buffer RAM 72 is specifically carried out according to the flowchart of FIG.

First, when data registration in the buffer RAM 72 starts, the address i of the buffer RAM 72 is set to "0" (S601). 18 bytes of data (one set of data) from the i address of the buffer RAM 72
The sum a of the square of the distance between the histogram data and the image quality data stored in the pRAM 76 is calculated according to the equation (10) (S602).

If the histogram data is normalized so as to take a value from 0 to 1, the total sum a is a predetermined reference value for judging the similarity, for example, 3.6 (the maximum value of the total sum a. (20% of 18)) is determined (S603). The value of 20% is an example, and the above-mentioned reference value is set arbitrarily according to the case.

If the total sum a is smaller than 3.6 in S603, both data are regarded as similar data, and the processing ends without registering the input data in the buffer RAM 72.
If the total sum a is 3.6 or more in S603, 18 is added to i to calculate the square of the distance to the next data in the buffer RAM 72 (S604).

When the capacity of the buffer RAM 72 is, for example, 100 sets of data, i and 18 × 100 are compared in size (S605). In 605, i is 18 × 1
If it is smaller than 00, that is, if the sum a of the data in the TempRAM 76 and all the 100 sets of data in the buffer RAM 72 is not calculated, the process returns to S602.
If i is 18 × 100 or more, then TempRAM
The data of 76 is registered in the buffer RAM 72 (S60
6), the process ends.

As described above, the similar data is stored in the buffer R.
By not registering in the AM 72, similar patterns are not learned, and more diverse patterns can be learned. Therefore, the image quality can be appropriately set for more types of original images.

[Fourth Processing System] This processing system, as shown in FIG. 25, includes a neuro / buffer unit 69, a switch 61.
71, basic neuro section 78, subtractor 79, adder 80
And a learning processing unit 82 having a switch 81. In this processing system, the histogram counters 52 ... 52
From the neuro / buffer unit 69 as a difference learning calculation unit via the switch 71.
Input to the basic neuro section 78.

Basic neuro section 7 as a basic learning operation section
Reference numeral 8 is a neuro computer that has learned in advance the correspondence between various image information and the image quality data most suitable for them, and the image quality data suitable for the input histogram data is calculated according to the learning content. Is configured to ask. The output side of the basic neuro section 78 is connected to the subtractor 79 and the adder 80.

The subtractor 79 as a subtraction unit is connected to the operation unit 3 and the switch 81. This switch 81
Is connected to an adder 80 as an adder and a switch 61, and the switch 61 and the subtractor 79 or the adder 8
The connection with 0 is switched. Switch 8
Specifically, 1 connects the switch 61 and the subtractor 79 during learning of the neuro computer 41, and connects the switch 61 and the adder 80 during copying in the automatic mode.

On the other hand, the adder 80 is connected to the internal parameter 59 via the switch 83.
The switch 83 is also connected to the operation unit 3, and switches the connection between the internal parameter 59 and the operation unit 3 or the adder 80. Specifically, the switch 83 connects the internal parameter 59 and the operation unit 3 during learning of the neuro computer 41, and connects the internal parameter 59 and the adder 80 during copying in the automatic mode.

The operation at the time of learning of the fourth processing system configured as described above will be described.

First, when the ID code number is registered by input from the operation section 3, one neuro / buffer section 69 is assigned to the ID code number, and the switch 61 is switched to the neuro / buffer section 69. To be During learning, the switch 81 connects the subtractor 79 and the switch 61, and the switch 83 connects the operation unit 3 and the internal parameter unit 59.

In this state, in the basic neuro section 78, 1 input from the histogram counters 52 ... 52 is input.
The image quality data is calculated based on the five pieces of histogram data. Then, in the subtractor 79, the image quality data output from the operation unit 3 by the user's input and the basic neuro unit 78
The difference from the image quality data output from is calculated.

The calculation result is input to the output side of the neuro / buffer section 69 selected through the switch 81. In addition, the neuro / buffer unit 69 has
The above histogram data is input through the switch 71. In the neuro / buffer unit 69, the histogram data and the above calculation result are used as one set of data and learning is performed in the same manner as the learning performed in the third processing system.

The internal parameters are determined by the internal parameter section 59 based on the image quality data output from the operation section 3. Then, when the process control unit 63 determines the process control value by the internal parameter, the process control value is sent to the process control unit 6.

When copying in the automatic mode after learning as described above, the switch 81 connects the neuro / buffer section 69 and the adder 80, and the switch 83 connects the adder 80 and the internal parameter section 59. . First, the histogram data from the histogram counters 52 ... 52 is input to the basic neuro section 78 and the neuro / buffer section 69 selected by the ID personal identification number. At this time, in the neuro / buffer unit 69, the switch 73
The contact point 73a and the contact point 73c are connected by the switch 74
The contact point 74c is connected to the contact point 74a.

As a result, the basic neuro section 78 and the neuro / buffer section 69 calculate the image quality data according to the learning content based on the input histogram data. These image quality data are added by the adder 80 and then given to the internal parameter section 59 through the switch 83. The internal parameter section 59 determines internal parameters based on the output of the adder 80. And
Copying is performed based on the internal parameters.

In the above learning, all the connection weights other than the threshold value W20k in FIG. 5B are set to 0 among the connection weights and the initial values of the threshold values of the neurocomputer 41. Further, the threshold value W20k is set to an appropriate positive value (for example, a value close to +6 in x in FIG. 6). With this value, the neuro computer 41 can be calculated by the formula (2).
When the output of is calculated, the output becomes almost zero. Therefore, even in the unlearned state, inappropriate image quality data will not be output.

As described above, in the fourth processing system, the neuro computer 41 learns the difference between the basic image quality data determined in advance by the basic neuro unit 78 and the image quality data determined by the user, and makes a copy. Sometimes, the basic neuro section 7
8 and the output of the neuro computer 41 are added and given to the internal parameter section 59.

In the control system which is not configured as described above, when the number of learnings is small immediately after installing the copying machine,
Due to the small amount of data, it is not possible to copy with good image quality. In addition, if the amount of data is small, there is a problem that learning does not converge when a plurality of image quality is set for the same document or a large number of similar data are input. However, in the fourth processing system, learning is performed in advance based on the basic image quality data, and learning is further repeated according to the user's preference. It is possible to copy with various characteristics.

<Modification> As shown in FIG. 26, a modification of the present processing system basically comprises a fourth processing system as a basic configuration, and further, the learning content of the neuro computer 41 is I.
It is configured so that it is erased for each D personal identification number. The learning contents described above are stored in the connection weight memories 69a provided in each neuro / buffer unit 69. In the present processing system, for example, using the numeric keypad 38 on the operation unit 3, I
By inputting the D number, the coupling load in the coupling load memory 69a of the neuro / buffer unit 69 of the ID secret code is erased.

Specifically, among the initial values of the connection weight and the threshold value of the neuro computer 41, FIG.
All coupling loads other than the threshold value W20k are set to zero. In addition, the threshold value W20k is set to an appropriate positive value (for example,
(Value close to +6 in x in FIG. 6). This allows
When the image quality condition is set, the output of the neuro computer 41 becomes almost zero.

In the control system configured as described above, I
When changing the D personal identification number, first, the learning content of the neuro computer 41 corresponding to the ID personal identification number to be changed is erased. This is executed by the user's input operation using the operation unit 3. The number for setting the mode for erasing the ID code number is predetermined, and the erasing mode is entered by inputting it with the ten keys 38 on the operation unit 3. Then, if the ID code number is similarly input with the ten-key pad 38, the learning content of the neurocomputer 41 corresponding to the ID code number is erased. In this state, a new ID personal identification number corresponding to the neuro computer 41 is input from the operation unit 3 and registered.

In this way, it is possible to easily re-learn the neuro computer 41. For example, when some problem occurs in the learning content, unless the learning content is deleted, the copy image quality as set cannot be obtained even if new learning is repeated. In addition, when a user who has been using a particular neurocomputer 41 gives the right to use the neurocomputer 41 to another user, if the learning content of the previous user remains, the user who will use it later learns newly. Cannot be performed and the image quality desired by the user cannot be obtained. However, if re-learning can be easily performed as in the present processing system, unintentional image quality setting will not be performed due to unnecessary learning content.

[Fifth Processing System] The present processing system includes a neuro computer 84 shown in FIG. The neuro computer 84 includes a ROM 75, a CPU 77,
.., a switch 86, and a CPU bus 87.

The CPU 77 is adapted to perform a neuro operation according to the program stored in the ROM 75 as described above, and constitutes a processing unit together with the ROM. In the present processing system, the CPU 77 further switches the switch 86 as the connection switching means in accordance with the ID code output from the operation unit 3 to connect the coupling load memory 85 corresponding to the ID code to the CPU bus 87. It has a function to do.

The connection weight memory 8 as a learning content storage unit
.. are provided for each ID personal identification number, and the coupling weight as the learning content is stored in each. As a result, W1 [i] [j] in FIGS. 7, 8 and 9 becomes W1 [ID] [i] [j], and W2 [j] [k] becomes W2 [ID] [j] [k]. Becomes

As described above, in this processing system, a plurality of coupling load memories 85 ... Are provided separately from the CPU 77, and are selected by the ID code number and connected to the CPU 77. Therefore, according to this processing system, the neuro-calculation unit centered on the CPU 77 is
D It is not necessary to provide multiple numbers according to the PIN code, and only one is required. Therefore, the neuro computer 84
The configuration of is simplified.

<Embodiment 2> Next, another embodiment of the present invention will be described with reference to FIG.

In this embodiment, as shown in FIG. 28, a plurality of copying machines 91 to 94 are provided, and these are network 95.
An example in which the system is configured by being connected to each other by means of the above will be described.

Each of the copying machines 91 to 94 is provided with neuro computer parts 91a to 94a. The neuro computer units 91a to 94a are the learning processing units 58 and 70 described above.
As indicated by 82, the relationship between the image information and the image quality condition is learned, and an appropriate image quality condition is calculated based on the learning content. The copying machine 91 functions as a master machine and stores the learning content. Also, the copying machine 92-
The reference numeral 94 functions as a child device, and temporarily stores the learning content.

When copying is performed by the copying machine 91, it is performed in the same manner as in the first embodiment. When copying is performed by the copying machines 92 to 94 as external devices, the copying machine 9 is used.
2 to 94 send the ID personal identification number to the copying machine 91, receive the learning content corresponding to the ID personal identification number from the copying machine 91, and set the image quality according to the learning content. Also,
The learning contents corrected by the learning in the copying machines 92 to 94 are sent back to the copying machine 91.

With this structure, it is not necessary to use the memory cards 60 and 65 in the first embodiment. Further, any of the copying machines 91 to 94 can perform copying using the same learning content, and the user does not need to remember which copying machine 91 to 94 stores the learning content to be used.

[0251]

As described above, in the image forming apparatus for setting the image quality by the neural network according to the first aspect of the present invention, the image information reading means for reading the image information of the original image and the image information reading means for reading the image information. Based on the image forming means for forming a copy image on the transfer material based on the obtained image information, the image quality condition input means for inputting the image quality condition of the copy image, and the image quality condition input from the image quality condition input means. A control unit for controlling the image forming unit, the image information read by the image information reading unit as input data, the image quality condition input from the image quality condition input unit as teacher data, and the input data and the teacher data. The relationship is learned by associating it with the identification number, and according to the learning content about the identification number specified when copying the original image. It is configured to have a quality condition determining means having a neuro-computer for calculating the image quality condition suitable for the image information read by the image information reading means.

Thus, the relationship between the image information and the image quality condition is learned, and the image quality condition is set according to the content of the learning. Therefore, it suffices to store the connection weight of the neurocomputer as the learning content, and it is not necessary to provide a large-capacity memory as in the configuration for storing the above relationship in the memory or the like. Moreover, by learning and setting for each identification number, if the user uses a unique identification number, the image quality can be set for each user.

Therefore, if the image forming apparatus according to the first aspect is adopted, it is possible to easily and automatically set the image quality according to the preference of each user with a simple structure.

The image forming apparatus for setting the image quality by the neural network according to claim 2 of the present invention is the image forming apparatus for setting the image quality by the neural network according to claim 1, which is from the image quality condition inputting means. A conversion unit that converts the input image quality condition into a parameter suitable for controlling the image formation, and a control value setting unit that sets a control value for controlling the image formation unit based on the parameter from the conversion unit are provided. The control means controls the image forming means based on the control value set by the control value setting means, and further uses the image information read by the image information reading means as input data,
The parameter from the converting means is used as the teacher data to learn the relationship between the input data and the teacher data for each identification number, and when the identification number is specified when copying the original image, the identification number is learned. According to the contents, there is provided a parameter determining means having a neuro computer for calculating a parameter suitable for the image information read by the image information reading means.

Thus, the relationship between the image information and the parameter is learned, and the image quality condition is set according to the content of the learning. Therefore, similar to the image forming apparatus according to the first aspect, it is not necessary to provide a large-capacity memory, and the user uses the unique identification number to set the image quality for each user. You can

Therefore, if the image forming apparatus according to the second aspect is adopted, it is possible to easily and automatically set the image quality according to the preference of each user with a simple structure.

The image forming apparatus for setting the image quality by the neural network according to claim 3 is the image forming apparatus according to claim 1 or 2, wherein the image quality condition determining means or the parameter determining means performs learning and learning. A processing unit that performs an operation according to the content, and a plurality of learning content storage units that store the learning content and that are provided for each recognition number;
Since the learning content storage unit of the designated identification number is provided with the connection switching means for connecting to the processing unit, the learning content can be saved for each identification number, and one processing unit is provided. You just have to provide it.

Therefore, if the image forming apparatus according to the third aspect is adopted, the structure of the image quality condition determining means or the parameter determining means can be simplified, and the cost of the image forming apparatus can be reduced. Play.

The image forming apparatus for setting the image quality by the neural network according to claim 4 is the image forming apparatus according to claim 1 or 2, wherein the image quality condition determining means or the parameter determining means is an external storage medium. Since the learning content is stored in the image forming apparatus, the learning content can be used in the image forming apparatus by mounting the external storage medium in another image forming apparatus.

Therefore, if the image forming apparatus according to the fourth aspect is adopted, the learning contents can be effectively used,
The effect is that the efficiency of image quality setting can be improved.

The image forming apparatus for setting the image quality by the neural network according to claim 5 is the image forming apparatus according to claim 4, wherein the image quality condition determining means or the parameter determining means includes a plurality of the external storages. Since the learning content for each recognition number is stored for each medium, if the recognition number is assigned to each user, the learning content can be used for each user in another image forming apparatus.

Therefore, if the image forming apparatus according to the fifth aspect is adopted, it is possible to easily manage the learning content for each recognition number when the learning content is shared by other image forming apparatuses.

The image forming apparatus for setting the image quality by the neural network according to claim 6 is the image forming apparatus according to claim 1 or 2, wherein the image quality condition determining means or the parameter determining means is used for learning. Since a configuration is provided in which a data storage unit that stores a pair of input data and teacher data as learning data is provided and learning is performed based on the learning data stored in this data storage unit, learning is performed based on many learning data. You can therefore,
The image quality setting is not specialized only for the image information of a specific document.

Therefore, if the image forming apparatus according to the sixth aspect is adopted, there is an effect that more accurate image quality setting can be performed as compared with the case where learning is performed with a small amount of learning data.

The image forming apparatus for setting the image quality by the neural network according to claim 7 is the image forming apparatus according to claim 6, wherein the image quality condition determining means or the parameter determining means is used in the data storage section. Since learning is performed after a certain amount of learning data is accumulated, biased learning is not performed when the amount of learning data is small at the beginning of use of the image forming apparatus.

Therefore, if the image forming apparatus according to the seventh aspect is adopted, it is possible to prevent the learning delay caused by the small amount of learning data.

The image forming apparatus for setting the image quality by the neural network according to the eighth aspect is the image forming apparatus according to the sixth aspect, wherein the image quality condition determining means or the parameter determining means is the last for the same document. Since only the input learning data is stored in the data storage unit, a plurality of learning data will not be adopted for the same document.

Therefore, when the image forming apparatus according to the eighth aspect is adopted, it is possible to prevent the inconvenience that the learning of the neurocomputer is not converged and to always perform the learning with the appropriate learning data.

The image forming apparatus for setting the image quality by the neural network according to the ninth aspect is the image forming apparatus according to the sixth aspect, wherein the image quality condition determining means or the parameter determining means receives the input learning data. When it is determined that the learning data is similar to the learning data already stored in the data storage unit, the learning data is excluded from the target to be stored in the data storage unit. Therefore, the learning of different patterns is performed more.

Therefore, if the image forming apparatus according to the ninth aspect is adopted, it is possible to prevent the similar learning data from being redundantly learned, and to perform the learning better and efficiently. Play.

The image forming apparatus for setting the image quality by the neural network according to the tenth aspect is the above-mentioned first aspect.
Alternatively, in the image forming apparatus according to the second aspect, the image quality condition determining unit or the parameter determining unit has learned the relationship between the standard image information and the image quality condition or the control value in advance.
A basic learning calculation unit that calculates the image quality condition according to the learning content, a subtraction unit that calculates the difference between the output value of this basic learning calculation unit and the teacher data, and the output of this subtraction unit as difference teacher data And a difference learning data, and a difference learning calculation unit that calculates a difference between the image quality condition and the output value of the basic learning calculation unit according to the learning content, and an output value of the difference learning calculation unit and the basic value. An addition unit that calculates the sum with the output value of the learning calculation unit is provided.

As a result, the value obtained as a result of the learning is the basic image quality condition plus the value learned according to the user's preference. Therefore, since the basic image quality is ensured, the image quality to be set gradually becomes closer to the user's taste by repeating learning.

Therefore, if the image forming apparatus according to the tenth aspect is adopted, it is possible to achieve stable image quality setting by performing basic image quality setting even when the learning number is small.

The image forming apparatus for setting the image quality by the neural network according to claim 11 is the above-mentioned claim 1.
In the image forming apparatus according to No. 0, the image quality condition determining unit or the parameter determining unit is configured to erase the learning content of the difference learning calculation unit for each recognition number, and thus the learning content corresponding to a specific identification number. Only can be erased.

Therefore, if the image forming apparatus according to the eleventh aspect is adopted, it is possible to easily change the learning content for each recognition number.

The image forming apparatus for setting the image quality by the neural network according to the twelfth aspect is the image quality forming method according to the first, second, third, fourth, fifth, sixth, seventh, eighth, ninth or tenth aspect. An apparatus, which is connected to an external device having a function equivalent to that of the image quality condition determining unit or the parameter determining unit, so that the contents learned by the image quality condition determining unit or the parameter determining unit can be shared with each other. Therefore, the same learning content can be used in either the image forming apparatus or the external device. In particular, if an image forming apparatus having the same function as this image forming apparatus is prepared as an external device, for example, the learning content stored in the image forming apparatus serving as the parent machine is called to the image forming apparatus serving as the slave machine. It is possible to use it and to return the learning contents modified by the child machine to the parent machine.

Therefore, if the image forming apparatus according to the twelfth aspect is adopted, the learning or the learning contents can be executed by any of the plurality of connected image quality forming apparatuses, and the image quality setting based on the learning can be performed more efficiently. There is an effect that it can be performed.

[Brief description of drawings]

FIG. 1 is a block diagram showing a first processing system of a copying machine according to an embodiment of the present invention.

FIG. 2 is a front view showing the internal structure of the copying machine.

FIG. 3 is an explanatory diagram showing an operation unit in the above copying machine.

FIG. 4 is an explanatory diagram showing a configuration of a neuro computer.

FIG. 5 is an explanatory diagram showing a configuration of units in an intermediate layer and an output layer of the neuro computer.

FIG. 6 is a graph showing a sigmoid function used to calculate each output value of an intermediate layer and an output layer in the neuro computer.

FIG. 7 is a flowchart showing a processing procedure at the time of learning of the neuro computer.

FIG. 8 is a flowchart following the flowchart of FIG. 7, showing a processing procedure at the time of learning of the neuro computer.

FIG. 9 is a flowchart showing a processing procedure when the image quality of the neuro computer is set.

FIG. 10 is a graph illustrating a processing method of a histogram counter provided in the first processing system.

FIG. 11 is a block diagram showing a configuration of the histogram counter.

FIG. 12 is a flowchart showing a procedure of a copying operation after learning by the first processing system.

FIG. 13 is a graph showing an exposure amount with respect to an average value of densities of a certain color of a document.

FIG. 14 is a block diagram showing a configuration of a modification of the first processing system.

FIG. 15 is a block diagram showing a configuration of a second processing system of the copying machine according to the embodiment of the present invention.

FIG. 16 is a flowchart showing a procedure of a copying operation after learning of the second processing system.

FIG. 17 is a block diagram showing the configuration of a third processing system of the copying machine according to the embodiment of the present invention.

FIG. 18 is a block diagram showing a schematic configuration of a neuro / buffer section in the third processing system.

FIG. 19 is a block diagram showing a configuration of a neuro computer of a neuro / buffer section in the third processing system.

FIG. 20 is an explanatory diagram schematically showing a storage area of a buffer RAM in the neuro / buffer unit.

FIG. 21 is a flowchart showing a procedure of a copying operation after learning of the third processing system.

FIG. 22 is a flowchart showing a procedure of a copy operation of the first modification of the third processing system.

FIG. 23 is a flowchart showing a control procedure for writing data in the buffer RAM in the second modification of the third processing system.

FIG. 24 is a flowchart showing a control procedure for writing data in a buffer RAM in a third modification of the third processing system.

FIG. 25 is a block diagram showing the configuration of a fourth processing system of the copying machine according to the embodiment of the present invention.

FIG. 26 is a block diagram showing a configuration of a modified example of the fourth processing system.

FIG. 27 is a block diagram showing the configuration of a neurocomputer in a fifth processing system of the copying machine according to an embodiment of the present invention.

FIG. 28 is an explanatory view showing another embodiment of the present invention and showing a configuration in which a plurality of copying machines are connected via a network.

[Explanation of symbols]

3 Operation Unit (Image Quality Condition Input Means) 6 Copy Process Unit (Image Forming Means) 41 Neurocomputer 57 Image Information Reading Unit (Image Information Reading Means) 58/70/82 Learning Processing Unit (Image Quality Condition Determining Means, Parameter Determining Means) 59 internal parameter unit (converting unit) 60/65 card memory (external storage medium) 63 process control unit (control unit) 69 neuro / buffer unit (difference learning calculation unit) 69a coupling weight memory 72 buffer RAM (data storage unit) 75 ROM 77 CPU 78 Basic neuro part (basic learning calculation part) 79 Subtractor (subtraction part) 80 Adder (addition part) 85 Coupling weight memory (learning content storage part) 86 Switch (connection switching means) 92-94 Copier ( External device) 95 Network

Claims (12)

[Claims] 1. An image information reading means for reading image information of a document image, an image forming means for forming a copied image on a transfer material based on the image information read by the image information reading means, and an image quality condition of the copied image. Image quality condition inputting means for inputting the image information, control means for controlling the image forming means based on the image quality condition input from the image quality condition inputting means, and image information read by the image information reading means as input data. With the image quality condition input from the image quality condition input means as the teacher data, the relationship between the input data and the teacher data is learned by associating with the identification number, and the identification number specified when the original image is copied. A neuro computer for calculating an image quality condition suitable for the image information read by the image information reading means in accordance with the learning content of An image forming apparatus for setting image quality by a neural network, the image forming apparatus including: 2. A conversion means for converting the image quality condition input from the image quality condition input means into a parameter suitable for controlling image formation, and a control for controlling the image forming means based on the parameter from the conversion means. An image forming apparatus, comprising: a control value setting means for setting a value, wherein the control means controls the image forming means based on the control value set by the control value setting means. By using the image information read by the input data as input data and the parameters from the conversion means as teacher data, the relationship between the input data and the teacher data is learned for each identification number, and the identification number is used when copying the original image. When specified, a parameter suitable for the image information read by the image information reading means is set according to the learning content of the identification number. Image forming apparatus that sets an image quality by the neural network according to claim 1, characterized in that it comprises a parameter determining means having a neuro-computer for output. 3. The image quality condition determining means or the parameter determining means comprises: a processing section for performing learning and a calculation according to the learning content; and a plurality of learning content storage sections for storing the learning content and provided for each identification number. The image forming apparatus for setting the image quality by the neural network according to claim 1 or 2, further comprising: a connection switching unit that connects a learning content storage unit of a designated identification number to the processing unit. 4. The image forming apparatus for setting image quality by a neural network according to claim 1, wherein the image quality condition determining means or the parameter determining means stores learning contents in an external storage medium. 5. The image quality by the neural network according to claim 4, wherein the image quality condition determining means or the parameter determining means stores learning contents for each recognition number in each of the plurality of external storage media. Image forming device to be set. 6. The image quality condition determining means or the parameter determining means includes a data accumulating portion for accumulating a pair of input data used for learning and teacher data as learning data, and based on the learning data accumulated in the data accumulating portion. The image forming apparatus for setting the image quality by the neural network according to claim 1 or 2, wherein learning is performed by learning. 7. The image quality is set by the neural network according to claim 6, wherein the image quality condition determining means or the parameter determining means performs learning after a certain amount of learning data is accumulated in the data accumulating section. Image forming apparatus. 8. The image quality by the neural network according to claim 6, wherein the image quality condition determining means or the parameter determining means stores only the learning data input last for the same document in the data storage section. Image forming apparatus for setting. 9. If the image quality condition determining means or the parameter determining means determines that the input learning data is similar to the learning data already accumulated in the data accumulating section, the learning data is stored in the data accumulating means. The image forming apparatus for setting the image quality by the neural network according to claim 6, wherein the image forming apparatus excludes the image from the object to be stored in the unit. 10. A basic learning calculation unit that learns the relationship between standard image information and image quality conditions or control values in advance, and the image quality condition determining means or parameter determining means calculates the image quality conditions according to the contents of the learning. And a subtraction unit that calculates the difference between the output value of this basic learning calculation unit and the teacher data, and learn the relationship between the input data and the difference teacher data by using the output of this subtraction unit as difference teacher data A difference learning calculation unit that calculates the difference between the image quality condition and the output value of the basic learning calculation unit according to the content, and an addition that calculates the sum of the output value of this difference learning calculation unit and the output value of the basic learning calculation unit. An image forming apparatus for setting image quality by the neural network according to claim 1 or 2, further comprising: 11. The image formation for setting the image quality by the neural network according to claim 10, wherein the image quality condition determining means or the parameter determining means erases the learning content of the difference learning operation part for each recognition number. apparatus. 12. An external device having a function equivalent to that of the image quality condition determining means or the parameter determining means is connected to each other so that the contents learned by the image quality condition determining means or the parameter determining means can be shared. Claims 1, 2, 3, 4, 5, 6, 7, 8, 9, 10 characterized.
Alternatively, an image forming apparatus that sets the image quality by the neural network according to claim 11.