JPH06223232A - Processor for identifying/recognizing image pattern - Google Patents

Abstract

PURPOSE:To reduce the load of correcting work by identifying that a learned image pattern surely belongs to a correct character or graphic category in addition to the output of a character or graphic category most similar to an input image pattern. CONSTITUTION:A normalizing processing circuit 2 inputs a binary image pattern and applies the normalizing processing of a position and size to the image pattern, a high-bit integer preparing circuit 5 stores the pixel values of an image position out of the obtained image pattern in bit positions and prepares a high-bit integer based on a previously set bit position table, a low-bit integer preparing circuit 6 prepares a low-bit integer constituted of bits smaller than the number of bits constituting the high-bit integer, and an identifying/ recognizing circuit 7 judges whether at least one of the obtained integers is included in a section of an integer section table previously set up in a reference dictionary or not, identifies that the integer belongs to the category when the integer is included in the table, and if the integer is not included, sorts the integer based on a known identification function to display the identified result.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an identification processing device and a recognition processing device for a binary image pattern centering on characters and figures, and more particularly to many categories such as printed kanji, handwritten kanji, English letters, symbols and figures. The present invention relates to an identification / recognition processing device for images with many deformations.

[0002]

2. Description of the Related Art Traditionally, Kanji OCR (Optical Character)
In a character recognition processing device such as r Reader) or a drawing recognition device, a feature for recognition is extracted in the form of a vector from a character or a graphic pattern, and the feature pattern is compared with a standard pattern vector of each category in a standard dictionary created in advance. A method is known in which a discriminant function such as similarity or dissimilarity is obtained, and the most similar character or figure category is used as the recognition result.

[0003]

The standard dictionary provided in the conventional character recognition processing apparatus or figure recognition apparatus is generally created from a large amount of character / figure data. The number of categories, the frequency of appearance for each category, the degree of deformation that occurs in each category, etc. do not always match the data entered by the user, such as
There is a problem that the expected recognition performance cannot be obtained depending on the user.

Further, it is indispensable for the user to successively perform visual re-inspection and correction of erroneous recognition and rejected characters, and when a large amount of erroneous recognition and rejects occur, the user is burdened with a great burden. There was a problem that it took.

Further, how to correct an erroneously recognized input pattern so that it can be recognized thereafter is an essential problem of pattern recognition. At present, manufacturers of OCR and various recognition devices or device maintenance are required. The work of updating the dictionary and the method of creating the feature vector is performed by the trader, and the device is not always easy for the user to use.

As described above, the conventional recognition techniques such as OCR impose a heavy burden on the user in the correction work when an erroneous recognition occurs, and the erroneous recognition is erroneous. There has been a problem that means for correctly recognizing recognized characters and the like have not been sufficiently established thereafter.

The present invention has been made to solve the above problems, and an object of the present invention is not only to output a character or figure category most similar to an input image pattern by using a discriminant function, An object of the present invention is to provide a technique capable of surely identifying that the image pattern learned so far is the correct character or graphic category.

Another object of the present invention is to provide a technique capable of reducing the burden of correction work.

Another object of the present invention is to enable the recognition pattern to be registered in the standard dictionary by a simple means even if erroneous recognition occurs, and the recognition pattern can be accurately identified in the subsequent recognition. To provide the technology.

Another object of the present invention is to provide a technique capable of creating unlearned modified character patterns as needed and registering them in a standard dictionary or the like.

The above and other objects and novel features of the present invention will be apparent from the description of this specification and the accompanying drawings.

[0012]

In order to achieve the above object, in the first invention, a normalization process for inputting a binary image pattern and normalizing the position and size of the image pattern is performed. Means and the image pattern obtained by the normalization processing means, by referring to a binary bit position corresponding to each position of the image from a preset bit position table, the bit position of the image position A high bit integer creating means for storing a pixel value to create a high bit integer, and a bit less than the high bit integer for the high bit integer by using a preset equal frequency table and a pixel representative value table. A low bit integer creating means for creating a low bit integer, and at least one integer obtained by the high bit integer creating means and the low bit integer creating means. Is determined to be included in the section of the preset integer section table of the standard dictionary, and if included, it is identified as the category, and if not included in the integer section table, it is known. It is characterized in that it is provided with an identification / recognition means for performing classification by an identification function and an image input / output means for displaying the image information determined by the identification / recognition means.

According to the second invention, an image input / output means capable of inputting or displaying image information including at least a character or graphic category of the binary image pattern, and inputting the binary image pattern, a position with respect to the image pattern. And a normalization processing means for performing normalization processing of the size, and with respect to the image pattern obtained by the normalization processing means, refer to a binary bit position corresponding to each image position from a preset bit position table. A high bit integer creating means for creating a high bit integer by storing the pixel value of the image position in the bit position, and a preset equal frequency table and pixel representative value table for the high bit integer. A low-bit integer creating means for creating a low-bit integer composed of less bits than the high-bit integer, a high-bit integer creating means and a low-bit integer By using at least one or more integers obtained by the composition means, the integer section table of the standard dictionary corresponding to the character or figure category inputted by the image input / output means and the integer are inputted, and 0 is set for each integer bit. Alternatively, a unit for counting the frequency of 1 and updating the bit position table, the equal frequency table, the pixel representative value table, and the integer section table of the standard dictionary based on the frequency is provided.

In the third invention, at least one of the bit position table, the equal frequency table, the pixel representative value table, and the integer section table is used,
Displaying an unlearned binary image pattern and at least one high-bit integer or low-bit integer of the binary image pattern, and a binary image pattern created by the deformation pattern creating unit, An image input / output unit for inputting image information about the image pattern, the high bit integer or the low bit integer obtained by the deformation pattern creating unit, and an integer section table for the image information input by the image input / output unit are input. And count the frequency of 0 or 1 for each integer bit,
And a means for updating the bit position table, the equal frequency table, the pixel representative value table, and the integer section table of the standard dictionary based on the frequency.

[0015]

According to the first aspect of the present invention, an image pattern in which position and size are normalized is input to a binary image pattern, and the pixel value at each position of the input image is preset. Create a high-bit integer by storing it in the bit position indicated in the position table, and refer to the preset equal frequency table and pixel representative value table, and use a bit number smaller than the high-bit integer. A low bit integer to be constructed is created and, for example, the high bit integer and the low bit integer are compared with a preset standard dictionary to identify which character or graphic category the input image is. , Or recognizes, not only the character or figure category most similar to the input image pattern is output using the discriminant function, but also the image patterns learned up to that point are surely correct character or figure category. It can be identified or recognized that it is.

According to the second aspect of the present invention, the information of the binary image pattern not used for creating the standard dictionary is added, the integer section table of the standard dictionary is updated, and the information is used for creating the integer. Since the bit position table, the equal frequency table, and the pixel representative value table are updated, the burden of correction work can be reduced.

According to the third aspect of the invention, an unlearned binary image pattern (deformed pattern) is created by using the information in the standard dictionary, and an image of which character or figure category these deformed patterns belong to is created. By entering the information,
These modification patterns are added to update the integer section table of the standard dictionary, the bit position table used for creating the integer, the equal frequency table, and the pixel representative value table, so that the burden of the correction work can be reduced. Even if erroneous recognition occurs, the recognition pattern can be registered in the standard dictionary by a simple means, and the recognition pattern can be accurately identified or recognized in the subsequent recognition.

[0018]

Embodiments of the present invention will now be described in detail with reference to the drawings.

FIG. 1 is a block diagram showing the schematic arrangement of an image pattern identification / recognition processing apparatus according to the present invention. In FIG. 1, reference numeral 1 is a binary image pattern storage circuit and 2
Inputs a binary image pattern of N × N mesh (N is an integer) extracted from the binary image pattern storage circuit, and, for example, the position and size of A normalization processing circuit that performs normalization processing,
Reference numeral 3 denotes an image input / output device, which inputs information on classification results such as character or graphic category numbers, identification function values, and binary image patterns output from the identification processing circuit 4, displays the information, and externally displays the information. For example, the image information is input and displayed by a keyboard or a mouse.

Reference numeral 4 denotes an identification processing circuit which is a main part of the present invention, and the detailed configuration of the identification processing circuit 4 is shown in FIG.

In FIG. 2, reference numeral 5 denotes a high bit integer generation circuit, which is N × subjected to the normalization processing from the normalization processing circuit 2.
An N-mesh image pattern is input, and an integer of high bits is created by referring to the preset contents of the bit position table storage circuit 8.

Reference numeral 6 denotes a low bit integer creating circuit, which inputs the high bit integer created by the high bit integer creating circuit 5 and refers to the contents of the equal frequency table storage circuit 9 and pixel representative value table storage circuit 10 to Create a low-bit integer with fewer bits than a high-bit integer.

An identification / recognition circuit 7 inputs the integers created by the high bit integer creation circuit 5 and the low bit integer creation circuit 6, for example, and inputs them into the contents of the integer section table storage circuit 11 of the standard dictionary. It is determined whether or not it is included, and if it is included, the character or graphic category is output. If not included, recognition is performed using a known identification function, and the recognition result is output.

Reference numeral 8 is a bit position table storage circuit for storing a bit position table, 9 is an equal frequency table storage circuit for storing an equal frequency table, and 10 is a pixel representative value table storage circuit for storing a pixel representative value table.

Reference numeral 11 denotes an integer section table storage circuit,
An integer section table for storing integer sections created by the high bit integer creation circuit 5 and the low bit integer creation circuit 6 for each character category is stored.

A table updating circuit 12 inputs the image information output from the image input / output device 3, the contents of the integer section table of the integer section table storage circuit 11, and the high level input to the identification / recognition circuit 7. A bit integer or a low bit integer, or a high bit integer or a low bit integer obtained from the modified pattern creating circuit 13 is input, the number of 0 or 1 for each bit included in the input value is counted, and the count value is also calculated. Further, the contents of the bit position table storage circuit 8, the equal frequency table storage circuit 9, the pixel representative value table storage circuit 10 and the contents of the integer interval table storage circuit 11 are updated.

Reference numeral 13 is a deformation pattern creating circuit, which inputs the image information output from the image input / output device 3, and the contents of the integer section table of the integer section table storage circuit 11.
The contents of the bit position table storage circuit 8, the equal frequency table storage circuit 9, and the pixel representative value table storage circuit 10 are input to create a high bit integer or a low bit integer calculated by using the integer interval table for each category. ,further,
A binary image pattern is created using the bit position table or the equal frequency table and the pixel representative value table.

As a first operation example of the processing apparatus of this embodiment, the high bit integer creation circuit 5 and the low bit integer creation circuit 5 in the identification processing circuit 4 will be mainly described below with a focus on character pattern identification processing and recognition processing. 6, when only the identification / recognition circuit 7, the bit position table storage circuit 8, the equal frequency table storage circuit 9, the pixel representative value table storage circuit 10, and the integer section table storage circuit 11 operate, that is, the high bit integer creation circuit 5 The case where the normalized binarized character pattern obtained by the normalization processing circuit 2 is input and the input character is recognized or identified will be described in detail as an example.

As the first method, an example will be described in which a bit position table, an equal frequency table, a pixel representative value table, and an integer section table are provided for each character category.

First, the high bit integer creation circuit 5 is provided to the normalization character pattern X of the input character obtained by the normalization processing circuit 2.
Is entered. The normalized character pattern X is X = (x ₀ ,
x ₁ , ..., X _i , ..., X _n−1 ) can be represented by a vector having n (n = N × N) elements, and each element x _i is a character position i (= 0, 1, ..., n-1) pixel value (for example, x _i = 0)
Is a white point and x _i = 1 is a black point).

For this normalized character pattern X, the high bit integer creation circuit 5 outputs the bit position table storage circuit 8
Of the bit position table A _j = (a _0j , a _1j , ..., A _ij , ... A _n-1j ) corresponding to the j-th character category of H _j (n-bit binary number) To create.
That is, the high-bit integers H _j is created by storing the pixel values x _i of the character position i a _ij-th bit of the high bit integer H _j.

The low bit integer circuit 6 inputs the high bit integer H _j of the high bit integer circuit 5, and the equal frequency table E _j = (e _0j ,
e _1j , ..., E _kj , ... E _n-1j ), and the pixel representative value table F _j = (f _0j ,
f _1j , ..., F _kj , ... F _n-1j ) (f _kj represents the representative value of the kth bit of the high bit integer of the jth character category)
, The low bit integer L _{j of the jth} character category
Create (n-bit binary number).

In this creation, when T (T> 0) elements of the equal frequency table E _j have the same value,
The number of bits t at which the value h _k of the k-th bit of the high-bit integer H _j is equal to f _kj is obtained, and for example, the value l _k (the value of the e _kj-th bit of L _j is determined by the following determination. l is the lowercase letter for the letter ell).

If T / t _k > t, then l _k = 0, otherwise l _k = 1 (t _k is a preset threshold of t) The high bit integer H _j and low bit integer L Find _j for all character categories.

The identification / recognition circuit 7 refers to the integer section table P _j of the integer section table storage circuit 11 to output the high bit integer H _j , low bit integer L _j and integer section output from the low bit integer creation circuit 6. _Collate with table P _j . J th
The integer section table P _j of character category constituted the starting point of the interval p _m from the size s _m of the section m _j-number of pairs,
_{P j = {(p 1,} s 1), (p 2, s 2), ..., (p m,
s _m ), ..., (p _mj , s _mj )} are stored. Each p _m
Is an integer composed of n bits for the high bit integer H _{j and} (e _n-1 +1) bits for the low bit integer L _j .

An example of collating the high bit integer H _j with the corresponding integer interval table P _j will be shown. High bit integer H _j is the interval (p _m, s _m) be contained in, it is determined that could be identified that character category, and outputs. If it is not included in any of the m _j sections of the j-th character category,
The difference d _j between the integer of the input character and each section is obtained (the difference between the integer of the input character and each section is d _j ).

Next, the character category having the smallest d _j among all the character categories is output as the recognition result.

The identification method is the same as that for the high bit integer H _j when the low bit integer L _j and the corresponding integer interval table P _j are collated. If it cannot be identified, for example, the difference d between the integer of the input character and each section of the j-th character category is calculated by the same method as the method of obtaining d _j in the case of the high-bit integer.
ask for _j . Next, each d _j is calculated for all the character categories, and the character category with the smallest d _j is output as the recognition result.

Further, the high bit integer H _j and the low bit integer L _j
Against the integer section table P _j for each, either one even intervals (p _m, s _m) be contained in the identification process determines that could be identified that character category is also feasible.

Further, as a method of calculating the difference d _j between the high bit integer H _j or the low bit integer L _{j of} the input character and the integer interval table for each integer, it is possible to use a known discrimination function.

In the first method, the bit position table, the equal frequency table, the pixel representative value table,
I explained the example when you have an integer interval table,
The basic operation is the same when the bit position table, the equal frequency table, the pixel representative value table, and the integer interval table that are common to a plurality of character categories are set.

Next, as a second operation example, the identification / recognition circuit 7 in the image input / output device 3 and the identification processing circuit 4, the bit position table storage circuit 8, the equal frequency table storage circuit 9 centering on the character pattern, Pixel representative value table storage circuit 1
0, integer section table storage circuit 11 and table update circuit 12 only operate, that is, the determination processing circuit 7 identifies or recognizes the input character, outputs the classification result to the image input / output device 3, When the information of the character category to be added from the image input / output device 3 to register the unlearned character is input to the table updating circuit 12,
An example in which the table updating circuit 12 updates the bit position table, the equal frequency table, the pixel representative value table, and the integer section table will be described in detail.

First, the image input / output device 3 inputs information regarding a character category to be updated (referred to as a jth character category) to the table updating circuit 12. The table updating circuit 12 inputs the values of the high-bit integer H _j and the low-bit integer L _j of the identification / recognition circuit 7 and the contents of each integer section table P _j of the standard dictionary, and the number of 0 and 1 of each bit. Is counted. Although the high-bit integer H _j and the low-bit integer L _j differ in the number of constituent bits, the basic operation for counting is the same, so the case of the high-bit integer H _j will be described.

In P _j , an interval (p _m , s _m ), that is, s _m integers p _m , p _m +1, p _m +2 starting from p _m and continuing
..., for p _m + s _m -1, a k-th bit of value p of frequency q _mk is p _m the value of the k-th bit is 1, the remainder p when the p _m divided by 2 ^k _0, s Remainder s ₀ when _m is divided by 2 ^{k + 1}
Is calculated by the following formula. In the formula, [.] Represents the integer part of the calculation result of.

[0045]

[Equation 1]

[0046]

[Equation 2]

Further, for the jth character category, when K ′ high-bit integers H _j are input from the judgment processing circuit 7, the frequency of 1 of the k-th bit of these K ′ integers is q ′ _k. In other words, the frequency Q _k of 1 at the k-th bit and the number K of integers used for learning in the total of the integer interval table and the high bit integer H _j are calculated by the following equation.

[0048]

[Equation 3]

[0049]

[Equation 4]

As a result, the frequency of 0s of the kth bit is K
−Q _k .

Further, as a result of calculating the frequency R _k of the larger one (the one being 1 when they are equal) with the frequency of 0 and 1 for each bit and rearranging in the ascending order of R _k , the k-th bit before updating is the If all the integers included in the section of the integer section P _j before updating are stored in k'th place, for example, by storing the value of the k-th bit of the integer before updating in the k'-th bit of the new integer, A new integer interval P _j 'is created.

Further, each element of the bit position table A _j , the equal frequency table E _j , and the pixel representative value table F _j is updated in the same manner. However, in the case of the pixel representative value table, the pixel representative value f _kj of the kth bit before updating the jth character category is
If K−Q _k > Q _k, then f _kj = 0, otherwise f _kj = 1.

As described above, according to the example of the updating process as described above, the table updating circuit 12 causes the bit position table storage circuit 8, the equal frequency table storage circuit 9, the pixel representative value table storage circuit 10, and the integer section table 11 to have a specific character category. To update.

Next, as a third operation example, focusing on the character pattern, the bit position table storage circuit 8, the equal frequency table storage circuit 9, and the pixel representative value table storage in the image input / output device 3 and the identification processing circuit 4 are stored. When only the circuit 10, the integer interval table 11, the table update circuit 12, and the modification pattern creation circuit 13 operate, that is, the modification pattern creation circuit 13 creates an unlearned modification pattern, and the image input / output device 3 is required. A detailed description will be given by taking as an example the case where the bit position table, the equal frequency table, the pixel representative value table, and the integer interval table are updated by adding the modification pattern in accordance with the above.

First, image information such as a character category for creating a modified character pattern is input from the image input / output device 3 to the modified pattern creating circuit 13 based on the character information of the identification / recognition result.

First example of operation of deformation pattern creating circuit 13
The method of creating from a high bit integer is shown below. The transformation pattern creating circuit 13 inputs the integer section table p _j of the j-th character category, and for each section (p _m , s _m ), for example, two integers are converted into two transformation characters using a bit position table. Image input / output device 3 after converting to pattern X
To display. The jth character in the displayed deformed character pattern
Regarding the modified character pattern included in the character category, for example, the table updating circuit 12 uses the bit position table, the equal frequency table, the pixel representative value table,
Update each table of integer interval table.

Next, as a second method of the operation example of the modified pattern creating circuit 13, a case of creating from a low bit integer will be shown. The transformation pattern creating circuit 13 inputs the integer interval table P _j for the low bit integer of the jth character category,
Wherein each section (p _m, s _m) with respect to, for example, two integer _{p m -α, p m + s} m -1 + α (α is an integer preset)
And compute these two integers in a bit position table,
It is converted into two modified character patterns X using the equal frequency table and the pixel representative value table and displayed on the image input / output device 3. In the case of this conversion, when each bit number of the integer L _j of the newly created modified character pattern corresponds to T (T> 0) positions of the binarized character pattern by referring to the equal frequency table. Is, for example, each bit of the integer L _j is 1
In the case of, each value of the corresponding pixel representative value table is assigned,
When it is 0, the complement of each value of the pixel representative value table is assigned. Among the displayed modified character patterns, the table updating circuit 12 is used to update each table of the bit position table, the equal frequency table, the pixel representative value table, and the integer section table for the modified character pattern included in the jth character category. To do.

In the first, second, and third operation examples, if there is a pattern that is not included in any character category, the table is created by using the pattern as a new character category. It is also possible to do so.

Further, in these operation examples, characters are mainly shown, but a general binary image such as a figure can be operated in the same manner.

As described above, according to the first, second and third operation examples, the high bit integer creation circuit 5, the low bit creation circuit 6, the identification / recognition circuit 7 and the bit position in the identification processing circuit 4 are described. Table storage circuit 8, equal frequency table storage circuit 9,
Pixel representative value table storage circuit 10 and integer section table 1
Since 1 is provided, not only the function of recognizing the input image pattern but also whether or not the input image pattern has been previously learned can be easily identified by referring to the integer interval table of the standard dictionary. For example, it is possible to store a large amount of image pattern information without increasing the storage capacity even if the number becomes large, such as tens of millions or trillions.

Even if an erroneous recognition occurs in the input character, the bit position table, the equal frequency table, the pixel representative value table, and the integer section table of the standard dictionary can be appropriately updated so that the erroneous recognition pattern can be identified. In addition, since it is possible to easily create deformation patterns, it is possible to easily increase unlearned data even for a category in which input image data is insufficient. Therefore, an identification / recognition processing device having high identification and recognition performance. Can be realized.

[0062]

As described above, according to the first invention, at least one integer obtained by the high bit integer creating means or the low bit creating means from the input image pattern is used, and the standard dictionary is used in advance. Since a means for identifying or recognizing the input image is provided by referring to the set integer section table, the OCR user can identify the erroneous recognition and the reject pattern sequentially by visual re-inspection and correction. It is not necessary to re-examine the captured image, and the work burden on the operator can be reduced.

Further, according to the second invention, even when the input image pattern is not included in the integer section table of the standard dictionary, that is, when a large amount of erroneous recognition or rejection occurs, for example, the bit position table is used. By providing the table updating means for updating the equal frequency table, the pixel representative value table, the integer interval table, etc., it is possible to reliably identify these erroneous recognition patterns thereafter. In other words, since the standard dictionary can be easily updated for each user, the more it is used, the more the recognition performance can be improved.

Further, according to the third aspect of the present invention, since an unlearned character pattern can be automatically generated based on the data of each table, it is possible to identify a character category to be newly registered or a character category registered so far. Even when it is desired to improve the recognition performance, it is possible to easily create an unlearned pattern and rewrite each table based on the information.

[Brief description of drawings]

FIG. 1 is a block configuration diagram showing a schematic configuration of an embodiment of an image pattern identification / recognition processing device of the present invention,

FIG. 2 is a block configuration diagram showing a configuration of an embodiment of an identification processing circuit which is a main part of the present invention.

[Explanation of symbols]

1 ... Binary image pattern storage circuit, 2 ... Normalization processing circuit,
3 ... Image input / output device, 4 ... Identification processing circuit, 5 ... High bit integer creation circuit, 6 ... Low bit integer creation circuit, 7 ... Identification
Recognition circuit, 8 ... Bit position table storage circuit, 9 ... Equal frequency table storage circuit, 10 ... Pixel representative value table storage circuit, 11 ... Integer section table storage circuit, 12 ... Table updating circuit, 13 ... Deformation pattern creating circuit.

Claims (3)

[Claims] 1. A normalization processing means for inputting a binary image pattern consisting of characters and figures and normalizing the position and size of the image pattern, and an image obtained by the normalization processing means. For a pattern, a high bit number is created by referring to a binary bit position corresponding to each image position from a preset bit position table and storing the pixel value of the image position at the bit position to create a high bit integer. Integer creating means and low bit integer creation for creating a low bit integer composed of less bits than the high bit integer by using a preset equal frequency table and a pixel representative value table for the high bit integer Means and at least one or more integers obtained by the high-bit integer creating means and the low-bit integer creating means, and the integer is a preset integer interval table in the standard dictionary. It is determined whether it is included in the table section, and if it is included, it is identified as a character or graphic category included in the section, and if it is not included, classification is performed by a known identification function. An identification / recognition processing device for an image pattern, comprising: a recognition means and an image input / output means for displaying the image information determined by the identification / recognition means. 2. An image input / output unit for inputting or displaying image information including at least a character or graphic category of a binary image pattern, and a binary image pattern which is inputted and displayed in a position and size with respect to the image pattern. With respect to a normalization processing unit that performs normalization processing and an image pattern obtained by the normalization processing unit, a binary number bit position corresponding to each image position is referred to from a preset bit position table, A high bit integer creating means for creating an integer by storing the pixel value of the image position at a bit position, and the high bit integer using the preset equal frequency table and pixel representative value table for the high bit integer A low bit integer creating means for creating a low bit integer consisting of less bits than the integer, and an integer obtained by the high bit integer creating means and the low bit integer creating means. Use at least one number,
The integer section table of the standard dictionary corresponding to the category of the character or figure input by the image input / output unit and the integer are input, the frequency of 0 or 1 is counted for each bit of the integer, and the frequency is calculated based on the frequency. And a means for updating the bit position table, the equal frequency table, the pixel representative value table, and the integer section table of the standard dictionary. 3. An unlearned binary image pattern and a high bit of the binary image pattern using at least one of the bit position table, the equal frequency table, the pixel representative value table, and the integer interval table. Deformation pattern creating means for creating at least one or more integers or low bit integers, and a binary image pattern created by the deformation pattern creating means, and image information including at least a category of characters or figures for the image pattern. Image input / output unit for inputting the high-bit integer or low-bit integer obtained by the modified pattern creating unit and an integer interval table for the image information input by the image input / output unit, The frequency of 0 or 1 is counted, and the bit position table and the equal frequency table are counted based on the frequency. 3. The image pattern identification / recognition processing apparatus according to claim 1, further comprising means for updating the table, the pixel representative value table, and the integer section table of the standard dictionary.