JPH0620098A - Character recognition device - Google Patents

Abstract

PURPOSE:To quickly perform the character recognition processing by collating characters with each standard character picture classified in groups each of which has the similarity higher than a preliminarily determined threshold. CONSTITUTION:A dictionary flag 16 as an 8X8-bit binary mesh flag, a dictionary 17 for rough classification as a file of standard patterns representing respective categories classified by rough classification, and a threshold 18 for each category which is used for detailed discrimination only in the case of the simple similarity of the category higher than the simple similarity obtained as the result of collation between the feature pattern of an inputted character picture and the dictionary 17 for rough classification are stored in a rough classification dictionary part 14. Groups of standard flag data matched to flag data are obtained, and collation with the standard character pictures of each group is performed, and collation with each of standard character pictures classified to groups whose similarities exceed the preliminarily determined threshold 18 is performed, and standard character pictures are outputted in the order of similarity as recognition results.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a character recognition apparatus for reading a document containing characters as an image and recognizing the read character image as a character.

[0002]

2. Description of the Related Art Conventionally, various recognition methods have been proposed for character recognition of characters including Chinese characters, but each Chinese character has a complicated pattern of one character, and even if the Chinese character type is only applicable Chinese characters. There are 1945 characters, and even in print type there are Mincho typeface and Gothic typeface. If handwritten characters are included, it is very difficult for the machine to recognize them. Therefore, attempts have been made to improve recognition accuracy and speed up processing by combining various methods such as a matching method and a structural analysis in each device.

As described above, since Kanji is complicated and has many character types, rather than recognizing one character at a time, it is roughly classified according to the characteristics and a group of similar characters (category). ), Then finely identify the group, and finally recognize it as one character.

[0004]

FIG. 6 is a flow chart showing a general operation of general classification which has been conventionally performed. When recognizing one character image cut out in step k1, feature extraction is performed in step k2, and step k3 is performed.
~ K5, for each category included in the large dictionary,
A simple similarity calculation is performed to compare the standard pattern representing each category with a numerical value to show how similar the standard pattern and the input character image are, and the result is obtained in step k6. All categories are rearranged in the order of similarity based on the numerical value of
The categories from the nth rank to the nth rank are set as the classification results of the major classification.

In steps k8 to k10, step k7
Performs detailed matching with all standard patterns in the character recognition dictionary included in the category output in, calculates the degree of similarity,
In step k11, the character codes are rearranged based on this degree of similarity, and in step k12, the first to mth highest character codes are output as the recognition result.

In such a case, since the simple similarity calculation and the rearrangement of the calculation results are performed even in the large classification, the calculation amount is large and the recognition processing requires a relatively long time.

An object of the present invention is to provide a character recognition device which reduces the amount of calculation in the recognition process and can perform high-speed character recognition process.

[0008]

According to the present invention, an input means for inputting a character image and a plurality of standard character images are classified into groups similar to each other, and standard flag data and a standard character image showing a characteristic for each group. The character dictionary in which is stored and the flag data indicating the characteristics of the input character image are obtained, the group of the standard flag data that matches this flag data is obtained, the standard character image of each group is collated, and the similarity is calculated in advance. A character recognition device characterized by including a recognition means for collating with each standard character image classified into a group equal to or larger than a predetermined threshold value and outputting as a recognition result in order of similarity.

[0009]

According to the present invention, the character recognition apparatus has a character dictionary in which a plurality of standard character images are classified into groups similar to each other, and standard flag data and a standard character image showing a characteristic are stored for each group. Then, the flag data indicating the characteristics of the character image input by the input means is obtained in the recognition means, a group of standard flag data that matches this flag data is obtained, and the standard character image of each group is collated to determine the similarity. Collates with each standard character image classified into a group equal to or larger than a predetermined threshold value, and outputs the recognition result in the order of similarity.

[0010]

1 is a block diagram showing the overall construction of an embodiment of the present invention. A CPU (central processing unit) 2 is a central processing unit for arithmetic processing that controls the entire apparatus. R
The OM (Read Only Memory) 3 is a memory that stores processing programs of the CPU 2. A RAM (random access memory) 4 provides a work buffer 22 for the CPU 2, an image memory 19 that stores coordinate data of an input image from an input unit 5 described later, a feature data memory 21 from the feature extraction unit 7, and a feature. The extraction unit 7 includes a large classification flag memory 20 created from the characteristic data.

The input unit 5 is a character image input device such as an image scanner, and outputs these character images to the image memory 19. The cutout unit 6 cuts out a character image for one character from the image stored in the image memory. CPU mentioned above
2 may do this.

The feature extraction unit 7 uses a predetermined method to
The feature amount of the cut-out character image is extracted, a recognition feature pattern to be described later is created, a large classification flag is created based on this, and the large classification flag memory 20 and the characteristic data memory 21 of the RAM 4 are respectively generated. Store.

The recognition section 8 comprises a large classification calculation section 12, a detailed identification section 13, a large classification dictionary section 14, and a recognition dictionary section 15. The large classification dictionary unit 14 receives a dictionary flag 16 that is a binary mesh flag of 8 × 8 bits, which will be described later, and a large classification dictionary 17 that is a standard pattern file representing each category classified according to the large classification. Compared with the simple similarity obtained as a result of collating the character image characteristic pattern with the large classification dictionary 17, detailed identification to be described later is performed only when the simple similarity of the category exceeds this value. Is a memory for storing the threshold value 18 for each category.

The large classification calculation unit 12 performs large classification based on the standard data stored in the large classification dictionary unit 14. The recognition dictionary unit 15 is a memory that stores a standard pattern created corresponding to each character, and the detailed identification unit 13 sets 1 for all characters in the category selected by the major classification.
Performs detailed matching with the standard pattern for each character. The interface 9 is an external output 1 such as a printer or a magnetic disk.
0 is a connection unit for outputting data from the character recognition device 1. The display unit 11 is a display device such as a liquid crystal display device, and displays the recognition result of the character recognition device 1 and the like.

FIG. 2 is a diagram for explaining the 8 × 8 mesh feature and the major classification flag which is an 8 × 8 bit binary mesh flag in the extracted feature pattern of the character image.

FIG. 2A is a diagram for explaining the 8 × 8 mesh feature. The 8 × 8 mesh feature means that one character image is divided into 8 meshes vertically and horizontally, and divided into 64 meshes.
Extract the feature quantity of the image in each mesh and
It is represented by an eight-dimensional vector or matrix.

FIG. 2 (2) shows a binary mesh flag created from the mesh features shown in FIG. 2 (1). The row in the mesh feature of FIG. 2 (1) corresponds to each 8 bits of the binary mesh flag, and the bit of the corresponding mesh flag is set to 1 at the position where the feature amount is maximum in each row of the mesh feature. , Other bits are set to 0.
For example, as a feature of a row (0, 11, 25, 3
6, 33, 20, 8, 7) is obtained, (00010000) is set in the flag for that row.

FIG. 3 is a diagram showing an example of 8 × 8 mesh feature amount extraction. First, cut out the character image 8 × 8
Of mesh. As shown in FIG. 3A, the character image is vertically and horizontally divided into eight equal parts. For example, as shown in FIG. 3B, the feature extraction unit 7 obtains a density distribution for each mesh divided into 64 equal parts. In general, the density distribution feature amount is obtained by dividing the number of black dots in the mesh by the total number of pixels forming the mesh and multiplying it by a constant to obtain an integer, in order to obtain a vertical and horizontal balance.

Focusing on the lower left mesh in FIG. 3A, the number of focal points in this mesh is 45, and the number of pixels in the horizontal direction is 8 (the number of dots). In FIG. 3A, in order to simplify the description of the mesh feature amount, the number of black dots is divided by the number of pixels in the horizontal direction of the mesh. here,
The density distribution feature amount is about 5. In this way, the feature amount for each mesh is extracted, and the mesh feature of FIG. 3C is obtained.

FIG. 4 is a diagram for explaining the large classification flag and the dictionary flag, which are 8 × 8 bit binary mesh flags.

FIG. 4 (1) is a large classification flag created from the mesh features of FIG. 3 (3). The bit of the major classification flag corresponding to the position where the feature amount of each line of the mesh feature shown in FIG. 3C is the maximum, for example, 1 of the mesh feature.
Row 4th column, 2nd row 4, 5th column, 3rd row 5th column, ..., 8th row 2,
Corresponding to the 7th column, each bit in the 1st row, 4th column, 2nd row 4, 5th column, 3rd row 5th column, ... Bit is set to 0.

FIG. 4B shows an example of the dictionary flag 16 preset for each category. In the dictionary flag 16, the bit corresponding to the position where the feature amount of each line may be maximized for each category is set to 1 and the bit which is not possible is set to 0.

When the large classification flag and the dictionary flag are matched with each other, if there is a bit in the large classification flag that is 1 and the corresponding bit in the dictionary flag is 0, the subsequent recognition processing is performed for that category. Do not do.

FIG. 5 is a flow chart showing the operation when the character recognition device 1 of this embodiment recognizes one character.

In step m1, the cut-out section 6 makes the image memory 1
A character image for one character is cut out from the image stored in 9, and is output to the feature extraction unit 7. In step m2, the feature extraction unit 7 creates a predetermined number of 8 × 8 mesh features by a predetermined method and outputs them to the feature data memory 21.

In step m3, the feature extraction unit 7 uses 8 × 8 bit 2 based on the 8 × 8 mesh feature.
A large classification flag that is a value mesh flag is created and stored in the large classification flag memory 20. In step m4,
In the recognition unit 8, the dictionary flag 16 of each category, the large classification dictionary 17, and the threshold value 18 from the large classification dictionary unit 14 are included.
Is read by the large classification calculation unit 12. In step m5, it is judged whether or not there is data to be collated in the large classification dictionary unit 14, and if there is data, the process proceeds to the next step m6.

In step m6, the large classification flag and the dictionary flag are collated. At this time, if there is a bit in the large classification flag that is 1 and the corresponding bit in the dictionary flag is 0, the subsequent processing is not performed for that category, the process returns to step m4, and the large classification processing of the next category is performed. I do. The category that matches the collation of the large classification flag and the dictionary flag is subjected to the processing of the next step m7.

In step m7, the large classification calculation unit 12
Compares the 8 × 8 mesh feature with the large classification dictionary 7 to calculate the simple similarity. At this time, the simple similarity is the inner product of the large classification dictionary feature vector and the 8 × 8 mesh feature vector.

In the next step m8, step m7
The value of the simple similarity calculated in, is compared with a threshold set in advance for each category, for the category of which the value of the simple similarity is less than or equal to the threshold, without performing the subsequent detailed identification, The process returns to step m4, and the process of the large classification of the next category starts. For categories whose simple similarity exceeds the threshold value, the process proceeds to the next step m9 to perform detailed identification.

In step m9, the recognition dictionary 15 is read to the detailed identification section 13, and detailed identification is performed for all the characters belonging to the category whose simple similarity exceeds the threshold value. In the detailed identification, the detailed identification unit 13 calculates the similarity between the 8 × 8 mesh feature vector and the recognition dictionary 15, and then returns to step m4 to read the next category in the large classification dictionary 7. When it is determined in the next step m5 that the collation with all the categories in the large classification dictionary 7 is completed, the process proceeds to step m10.

In the next step m10, step m
Based on the similarities calculated and accumulated in 9, the corresponding character codes are rearranged in the order of the similarities. In step m11, the first of the character codes rearranged in step m10
The rank to the nth rank are output to the RAM 3 as the recognition result.

As described above, in this embodiment, the binary flag determined in advance and the value determined by the obtained mesh feature are used.
The category for which the simple similarity is to be calculated can be limited by a simple calculation of matching with the value flag. Further, the category is not rearranged based on the value of the simple similarity, and the conformity or nonconformity is determined by comparing with a threshold value set in advance for each category. As a result, the calculation amount of the large classification can be reduced, and the character recognition processing can be speeded up.

[0033]

As described above, according to the present invention, the flag data indicating the characteristics of the input character image is obtained, the group of the standard flag data that matches this flag data is obtained, and the standard character image of each group is collated. Since the similarity is compared with each of the reference character images classified into groups whose similarity is greater than or equal to a predetermined threshold value, the amount of calculation is small in the recognition process of the group to which the input character image belongs, and therefore high-speed character recognition is performed. Processing can be performed.

[Brief description of drawings]

FIG. 1 is a block diagram showing the overall configuration of an embodiment of the present invention.

FIG. 2 shows 8 of the characteristic patterns of the extracted character image.
It is a figure explaining the large classification flag which is a * 8 mesh feature and a binary mesh flag of 8x8 bits.

FIG. 3 is a diagram illustrating an example of 8 × 8 mesh feature amount extraction.

FIG. 4 is a diagram illustrating a large classification flag which is a binary mesh flag of 8 × 8 bits and a dictionary flag.

FIG. 5 is a flowchart showing an operation when the character recognition device 1 of the present embodiment recognizes one character.

FIG. 6 is a flowchart showing a general classification operation which is conventionally performed.

[Explanation of symbols]

 1 character recognition device 2 CPU 5 input unit 7 feature extraction unit 8 recognition unit 12 large classification calculation unit 13 detailed classification unit 14 large classification dictionary unit 15 recognition dictionary unit 16 dictionary flag 17 large classification dictionary 18 threshold

Claims (1)

[Claims] 1. An input means for inputting a character image, a character dictionary in which a plurality of reference character images are classified into groups similar to each other, and standard flag data and a standard character image indicating the characteristics of each group are stored. Find the flag data that shows the characteristics of the input character image, find the group of standard flag data that matches this flag data, and compare with the standard character image of each group, and make a group whose similarity is greater than or equal to a predetermined threshold value. A character recognizing device, comprising: a recognizing unit that performs a collation with each of the classified standard character images and outputs the result as a recognition result in order of similarity.