JPH02300988A - character recognition device - Google Patents

Abstract

PURPOSE:To recognize a specified character such as '-' (Chinese character meaning unity), '1', '-', etc., at high speed and in addition, correctly by deciding an input character in which the value of the ratio of the number of character constituent elements to a character frame area is settled within a prescribed range to be the specified character. CONSTITUTION:Based on character picture information which is compand- processed so that width and height are settled at prescribed values by a width/ height extracting part 251 and a width/height comparing part 252, the ratio of the number of the character constituent elements measured by a number of character constituent element measuring part 254 to the area of the rectangular character frame with which the input character is in contact at top and bottom and right and left measured by a character frame calculating means 253 is obtained by a specified character discriminating part 255. Then, if the value of this ratio is settled within the prescribed range, the input character is decided to be the specified character such as '-', '1', '-', etc. Thus, the specified character can be recognized at high speed and in addition, correctly without executing recognizing processing by pattern matching, etc.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] This invention generally consists of a single line extending in the horizontal or vertical direction, such as r-J, IN, and "~". The present invention relates to a character recognition device that can accurately recognize specific characters.

2. Description of the Related Art An optical character reader (OCR) is known as a character recognition device that recognizes character information of a text by computer processing or the like. This OCR photoelectrically converts character information of characters to be recognized, such as handwritten characters and printed characters, and converts character image information based on the electrical signals obtained by this photoelectric conversion into one image.
Cut out each character. Then, in the recognition section, this 1
Input characters are recognized one by one according to a predetermined recognition algorithm based on character image information cut out character by character.

Conventionally, when recognizing an input character using the above-mentioned OCR, first the global characteristics of the character image information of the input character (
For example, major classification is performed based on complexity index, four-sided code, etc.), and several candidate characters are selected. Next, further detailed recognition is performed from among the candidate characters selected by the major classification based on detailed features (for example, composite similarity, mixed similarity, etc.) of the character image information.

<Problems to be Solved by the Invention> However, the conventional character recognition method in OCR mentioned above does not provide a classification algorithm or recognition algorithm for each character (category) in either major classification or detailed recognition. , all of the multiple characters to be classified or recognized must be classified or recognized according to the same algorithm. Therefore, by normalization process, r-J, rlj and "="
A specific character, such as , is also stretched to a predetermined width and height, resulting in a substantially rectangular shape filled with black information forming the character area. Therefore, the specific character may be erroneously recognized as another character with a large number of strokes (that is, with a large amount of black information forming the character area), resulting in a problem that the recognition rate of the specific character becomes low.

However, in order to increase the recognition rate of the above-mentioned specific characters, complex processing based on complex recognition algorithms is required.
A problem arises in that the cost of the character recognition device increases.

Therefore, an object of the present invention is to provide r-
An object of the present invention is to provide a character recognition device that can quickly and accurately recognize specific characters such as J, r 1 jJ-J, etc.

<Means for Solving the Problems> In order to achieve the above object, the present invention extracts the width and height of an input character by a width/height extraction unit according to input character image information, and extracts the width and height of the input character according to input character image information. Based on the width value and height value of the character, a width/height expansion/contraction unit performs expansion/contraction processing so that the width and height of the input character become predetermined values;
In a character recognition device that recognizes an input character based on the character image information that has been subjected to the stretching process, a rectangular character frame is formed in which the top and bottom and left and right sides of the input character are in contact, based on the character image information that has been subjected to the stretching process. a character frame area calculation unit that calculates the area of the character frame area, a character component number measurement unit that measures the number of elements constituting the character area of the input character based on the character image information subjected to the expansion/contraction processing; Based on the character frame area value from the frame area calculation section and the number of character components from the character component number measurement section, calculate the ratio between the number of character components and the character frame area, and calculate the number of character components. The present invention is characterized in that it includes a specific character determining section that determines, as a specific character, an input character in which the value of the ratio between the character frame area and the character frame area falls within a predetermined range.

Further, the character recognition device of the present invention may cause the specific character determination unit to determine whether a vertically long specific character or a horizontally long specific character based on the width value and height value of the input character extracted by the width/height extraction unit. It is desirable to distinguish between specific characters.

Function> When character image information is input, the width/height extractor extracts the width and height of the input character according to the input character image information, and extracts the width value and height of the extracted character. Based on the values of , the width/height expansion/contraction section performs expansion/contraction processing so that the width and height of the input character become predetermined values. Then, based on the character image information subjected to the expansion/contraction process, the area of a rectangular character frame where the upper and lower and left and right sides of the input character are in contact is calculated by the character frame area calculation unit. On the other hand, based on the character image information that has been subjected to the above expansion/contraction processing,
The number of elements constituting a character area of an input character is measured by a character component number measuring section.

Then, based on the value of the character frame area from the character frame area calculation section and the number of character components from the character component number measurement section, the ratio of the number of character components to the character frame area (specific character An input character that is determined by the determining section and for which the value of the ratio between the number of character constituent elements and the area of the character frame falls within a predetermined range is determined to be a specific character by the specific character determining section.Therefore, There is no need to perform recognition processing on input characters that are determined to be specific characters in this way.

The character recognition device also causes the specific character determination unit to identify a vertically long specific character and a horizontally long specific character based on the width value and height value of the input character extracted by the width/height extraction unit. If it is configured to distinguish between characters, it is possible to distinguish between a vertically long specific character and a horizontally long specific character and determine the specific character.

<Example> Hereinafter, the present invention will be explained in detail with reference to illustrated embodiments.

FIG. 1 is a block diagram of a character recognition device according to the present invention.

Character image information input from an input section 1 comprising an image scanner or the like is binarized by a preprocessing section 2 and subjected to noise removal processing. Furthermore, characters are cut out based on this binarized character image information from which noise has been removed, and after being subjected to processing such as normalization, it is output to the recognition unit 3. In the recognition unit 3, feature parameters of the character pattern are extracted according to the image information for each character from the preprocessing unit 2, and pattern matching is performed between the extracted feature parameters and a standard pattern to recognize the input character. It will be done. Then, the recognition result is displayed on the recognition result display section 4, and erroneous recognition or correction of the eject character is performed. The control section 5 has an input section 1. Pre-processing section 2. Control the recognition unit 3 and recognition result display unit 4 to execute character recognition operation.

FIG. 2 is a detailed block diagram of the preprocessing section 2 in FIG. 1. The communication section 21 is for exchanging information with the control section 5 of the OCR main body, and character image information input from the input section 1 and sent out from the control section 5 is taken in via the communication section 21. The image receiving section 22 is a buffer used when the noise removing section 23 and the normalizing section 24 perform image processing as described later. 22
is temporarily stored.

The noise removal section 23 removes noise such as dirt by erasing independent areas of several bits or less in the character image information.

Then, the character cutting unit 24 cuts out the character image information from which noise has been removed by the noise removing unit 23, character by character.

Thereafter, the normalization unit 25 performs normalization processing based on the character image information from which noise has been removed and extracted into individual characters, and variations in character size, slant position, line width, etc. are removed. Make the font size etc. constant. Furthermore, the normalization unit 25 executes specific character determination processing according to the present invention as will be described in detail later. Then, after predetermined processing is performed by the noise removal section 23 or the normalization section 25! The character image information for each character is sent to the control unit 5 via the communication unit 21.
sent to.

FIG. 3 is a more detailed block diagram of the part related to specific character determination processing in the normalization section 25.

Hereinafter, the determination process for specific characters such as r-1, rlj and "-" according to the present invention will be described in detail with reference to FIG.

The character image information taken in through the communication section 21, noise removed by the noise removal section 23, and stored in the image reception section 22 is read into the normalization section 25 and input to the width/height extraction section 251. be done.

The width/height extraction section 251 extracts the width and height of the character based on the input character image information for each character. Then, the width/height expansion/contraction section 252 becomes
Based on the width value and height value of the character from the width/height extraction unit 251, expansion/contraction processing is performed so that the width and height of the character become predetermined normalized reference values. At this time, the expansion/contraction process is performed, for example, as follows. That is, the character is expanded and contracted in the vertical and horizontal directions so that the upper and lower sides and left and right sides of the character touch a rectangular frame having a predetermined size.

In other words, in the case of a specific character rlJ that has been subjected to stretching processing, for example, it is stretched in the horizontal direction until it reaches a predetermined width, so that the rectangular frame is sufficiently filled with the black information that constitutes the character area. It is.

The width/height extraction section 251 and the width/height expansion/contraction section 252
The normalization process performed by is a normalization process that has been commonly performed in the past.

The character image information subjected to the expansion/contraction process by the width/height expansion/contraction section 252 is output to the character frame area calculation section 253 and the character component number measurement section 254.

The character frame area calculation unit 253 calculates the area W of the character frame of the minimum rectangle surrounding the character (that is, the rectangle in which the upper and lower sides and the left and right sides of the character are in contact) based on the input character image information.
Calculate. On the other hand, the character component number measuring unit 254 measures the number a of bits forming a character area based on the input character image information. ' Then, based on the value of the area W of the character frame from the character frame area calculation unit 253 and the value of the number of bits a forming the character area from the character component number measurement unit 254, the specific character determination unit 255 calculates the following: Calculate the determination constant C=a/w. This determination constant C represents the area of the character area relative to the area of the character frame, and the larger the value, the more the area of the character frame is approximately equal to the area of the character area, and the character frame has sufficient black information that constitutes the character area. It means being satisfied. That is, the judgment constant C
Character image information for which the value of is greater than a certain threshold value 0 is r
It can be said that specific characters such as -J, rlJ, and “=” are character image information expanded and contracted by the width/height expansion/contraction unit 252.

In this case, in the case of a character with a large number of strokes, even if the width/height expansion/contraction unit 252 performs expansion/contraction processing on the character image information, the character frame forms a black area that constitutes the character area, as in the case of the specific character described above. Since it is not filled with sufficient information and the value of the determination constant C is relatively small, it is easy to distinguish it from the above-mentioned specific character. Therefore, when the value of the determination constant C is larger than the threshold θ, the specific character determination unit 255 determines that the recognition category of the character is r−jJlj and “
It is determined that the character is a specific character such as "-".

Further, the specific character determining unit 255 determines whether the specific character determined as described above is based on the width value and height value of the character extracted by the width/height extracting unit 251.
Is it a vertically long specific character such as rlJ or r-J, r-
It identifies whether it is a horizontally long specific character such as J or "-".

In this way, the specific character determination unit 255 determines whether the input character is a vertically long specific character or a horizontally long specific character, and the determination result is transmitted to the character shown in FIG. 1 via the communication unit 21 in FIG. It is output to the control section 5. Then, there is no need to perform the recognition process for the specific character, so the control unit 5 instructs the recognition unit 3 not to perform the recognition process for the specific character. By doing this, recognition of specific characters can be made faster and
It can be done accurately.

At that time, if the specific character is written or printed with an inclination, there is a possibility that it will not be determined as a specific character depending on the above-described determination method. Therefore, if standard patterns for specific characters such as r-J, rlJ, and "-j are prepared in advance as in the case of other characters, the recognition unit 3 can recognize specific characters with an inclination. It is.

In this way, the character recognition device of this embodiment includes a width/height extraction section 251, a width/height expansion/contraction section 2521, a character frame and product calculation section 253, and a character component element in the normalization section 25 of the preprocessing section 2. It includes a measuring section 254 and a specific character determining section 255. Then, the width/height expansion/contraction unit 252 normalizes the width and height of the input character.

Then, a determination constant C( -a/w) is larger than the threshold θ, the specific character determination unit 255 determines the recognition categories of the character as r-J, rlJ, and “
It is determined that the character is a specific character such as "-". Furthermore, the above width/
Based on the width value and height value of the character extracted by the height extraction unit 251, the specific character determination unit 255 determines whether the character is vertically long or horizontally long, and the specific character is a vertically long specific character or a horizontally long specific character. Therefore, the specific character can be identified at high speed by simple processing without performing recognition processing by pattern matching performed in the recognition section 3. Furthermore, since the specific character is determined using the determination constant C based on the number of bits constituting the input character whose width and height have been normalized, the specific character can be reliably recognized.

In the above embodiment, the normalization unit 25 of the preprocessing unit 2 includes a character frame area calculation unit 2532, a character component measurement unit 254, and a specific character determination unit 255 related to specific character determination. However, the present invention is not limited to this (it may be provided independently of the normalization unit 25.
The recognition unit 3 may be provided to perform specific character determination processing as a preprocessing of recognition processing.

In the above embodiment, the recognition unit 3 recognizes the tilted specific character through normal recognition processing. However, the invention is not limited thereto. In other words, the threshold θ is smaller than the threshold θ above.
. and set this threshold value θ. (This includes specific characters that are tilted to some extent.) Then, the value of the determination constant C is θ. If it is within the range of ≦C≦θ, the normalization unit 25 performs rotation normalization on the character image information, and then calculates the determination constant C again.

As a result, when the value C' of the obtained determination constant C satisfies c'>e, the character may be determined to be the specific character.

<Effects of the Invention> As is clear from the above, the character recognition device of the present invention has the following effects:
The character frame area calculating section includes a side section for counting the number of character constituent elements and a specific character determining section, and the above-mentioned The ratio of the number of character constituent elements measured by the character constituent element number measuring unit to the area of a rectangular character frame where the top and bottom and left and right sides of the input character are in contact, calculated by the character frame area calculation unit, by the specific character determination unit. Find the input characters such that the value of this ratio falls within the predetermined range r-J, r! Since the characters are determined to be specific characters such as J and "-", the specific characters can be recognized quickly and accurately through simple processing without performing recognition processing such as pattern matching.

Further, the character recognition device of the present invention may cause the specific character determination unit to determine whether a vertically long specific character or a horizontally long specific character based on the width value and height value of the input character extracted by the width/height extraction unit. Since the specific character is identified from the specific character, the specific character can be determined by distinguishing between the vertically long specific character and the horizontally long specific character, and the specific character can be recognized more accurately.

[Brief explanation of drawings]

FIG. 1 is a block diagram of a character recognition device according to the present invention, FIG. 2 is a detailed block diagram of the preprocessing section in FIG. 1, and FIG. 3 is a block diagram of a specific character determination process of the normalization section in FIG. FIG. 3 is a more detailed block diagram of the parts; DESCRIPTION OF SYMBOLS 1... Input part, 2... Preprocessing part, 3... Recognition part, 4... Recognition result display part, 5... Control part, 21...
Communication unit, 22... Image receiving unit, 23... Noise removal unit, 24... Character extraction unit, 25... Normalization unit, 25
1...Width/height extraction section, 252...Width/height expansion/contraction section, 253...Character frame area calculation section, 254...Character component number measurement section, 255...Specific character determination Department.

Claims (2)

[Claims] (1) The width and height of the input character are extracted by the width/height extraction unit according to the input character image information, and the input character is extracted based on the width value and height value of the extracted character. In a character recognition device that performs expansion/contraction processing by a width/height expansion/contraction unit so that the width and height of Based on the processed character image information,
A character frame area calculation unit that calculates the area of a rectangular character frame where the upper and lower sides and left and right sides of the input character are in contact with each other, and the number of elements that make up the character area of the input character based on the character image information that has been subjected to the above expansion/contraction processing. based on the character frame area value from the character frame area calculation section and the character component number from the character component number measurement section,
A specific character determination unit that calculates the ratio between the number of character components and the area of the character frame, and determines an input character for which the value of the ratio between the number of character components and the area of the character frame falls within a predetermined range as a specific character. A character recognition device comprising: (2) The specific character determination section identifies vertically long specific characters and horizontally long specific characters based on the width value and height value of the input character extracted by the width/height extraction section. The character recognition device according to claim 1, characterized in that the character recognition device is configured as follows.