JPH07271911A - Character recognizing device - Google Patents

Abstract

PURPOSE:To provide a character recognizing device which can fast and accurately segment and recognize the characters and also can detect the margins regardless of a document printed in mixed full and half size characters or a document of unfixed pitches. CONSTITUTION:A character string extracting part 2 segments an input character string image, and a character element detecting part 3 extracts the character elements. A character width calculating part 4 calculates the widths of full and half size characters based on the height, etc., of the character images. A character element dividing part 5 detects and divides the character elements of widths less than the half size character width. A character candidate generating part 6 connects together the divided character elements to generate the full and half size character candidate images, and these candidate images are tentatively recognized by a recognizing part 7. The result of this recognition is evaluated again by a recognition result evaluating part 8. Thus the correct characters are recognized. Then a margin adding part 9 decides the presence or absence of inter-character margins based on the calculated character pitch.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a character recognition device,
In particular, the present invention relates to a character recognition device that recognizes a character in an input document image and converts it into a character code.

[0002]

2. Description of the Related Art In recent years, a printed document is read by optical / electrical conversion or the like, and once converted into image data information in which each pixel is converted into bit information, and then characters in the image data are recognized and data is recognized. Research and development of a system for reducing the input power, translating it into a foreign language, and uttering it for the blind, the visually impaired, children, and school children have been made, and some have been put into practical use.

The present invention relates to a character recognition device used in such a system. By the way, in the conventional character recognition device adopted in such a system, a black pixel is projected in a vertical direction on a character string image cut out, and a region of continuous orthographic projection is treated as a character image. Was cut out (for example, Akiyama et al. “Cutout of characters in article area of printed matter” PRL 80-70). By the way, regardless of whether the recognition target is a simple one such as an alphabet, in Japanese, Chinese, etc., there are many types and numbers of characters to be recognized. It is necessary to know exactly the shape and size of the characters that are displayed. In addition, even in printed documents in Western languages, various types of symbols are used in scientific and engineering papers, etc., so it is important to accurately determine the size of characters to be recognized. However, in the above-described character cutout technique, a character image is cut out using a region in which the orthogonal projection is continuous as a character. Therefore, when there is contact between characters (hereinafter referred to as contact characters), for example,
It is a separated character consisting of multiple character elements separated in the character string direction, such as "North" or "River" in a horizontal writing document or "2" or "3" in a vertical writing document, or a character originally consisting of a single character element. However, if there is a character that has been separated into a plurality of character elements due to faintness (hereinafter, both are collectively referred to as a separated character), accurate cutting cannot be performed.
Therefore, some methods have been proposed for correctly extracting characters from a character string image of a document including contact characters and separated characters. For example, in Japanese Unexamined Patent Publication No. 5-128308 “Character recognition device”, a character element having a width narrower than a predetermined character size is treated as a separated character, and a character element before and after (left, right, top and bottom) is regarded as a character element within a width of the character size. Connect and cut out as one character. Also, character elements wider than the specified character size are treated as contact characters, and these are divided by character size.
Each of these divided pieces is cut out as one character. When a character element that is narrower than the specified character size and a character element that is wider than the specified character size are consecutive, the separation position when dividing by character size from the beginning position of the narrow character element The cut-off candidate position is defined as the cut-off candidate position from the start position of the wide character element for each character size, and an image sandwiched between two different cut-out candidate positions among these cut-out candidate positions is once written as a character. Is temporarily recognized, and a character having a higher evaluation value is selected as the original character. This takes into consideration the possibility that the latter half element of the separated character is in contact with the next character.
On the other hand, in Japanese Unexamined Patent Publication No. 5-128307 “Character recognition device”, while Japanese Unexamined Patent Publication No. 5-128308 generates cut-out candidate positions for each fixed character size, half-width character width is applied to the character size. All the separation positions in the case of doing and the case of applying the full-width character width are all candidates.

Further, regarding the detection of the blank space in the character string,
The interval width of the cut out characters is compared with the full-width character width or the half-width character width, and when the margin width is larger, the margin is defined.

[0005]

However, the method disclosed in Japanese Patent Laid-Open No. 5-128308 is based on the premise that the character size is constant. For this reason, it is not possible to properly deal with a document printed in both full-width and half-width and a document with an inconstant pitch. Moreover, technical papers and the like have a large number of documents in these modes. On the other hand, in the method disclosed in Japanese Patent Laid-Open No. 5-128307, the number of cutout candidate positions is very large, and all the images sandwiched between two different cutout candidate positions among the cutout candidate positions are recognized as characters once. Since the processing is performed in the same manner, the amount of calculation becomes very large, and the processing speed of the entire system is delayed.

Further, in the detection of the blank space, when the full width character width is used as the character width, the narrow half width blank space cannot be detected. On the other hand, when the half-width character width is used as the character width, for example, when a narrow character continues in a document printed with a full-width pitch, the margin width between characters is a half-width character. Since the width is larger than the width, an extra margin is detected. As a result, it is difficult to properly deal with documents printed with full-width and half-width characters and documents with irregular pitch.

In view of the above problems, the present invention is a character recognition device capable of fast and accurate character cutting and margin detection even for a document printed with a mixture of full-width and half-width characters and a document with an irregular pitch. It is made for the purpose of providing.

[0008]

In order to achieve the above object, according to the invention of claim 1, a character string image is cut out from a document image input by converting bit information into pixels for each pixel, and then each character is extracted. In a character recognition device that further cuts out individual character images from a column image, recognizes each character image on this, and converts into a corresponding character code, a character element that is a group of pixels forming a character from the character string image. (Consisting of line segments and curves having the same phase with respect to the character string direction or a combination thereof), and a full-width character from the position or size of the character element extracted by the character element extraction unit. Character width and half-width character width calculation unit that calculates the width of a character by a predetermined means (including, procedure, method), and half-width calculated by the character width calculation unit among the character elements extracted by the character element extraction unit Sentence If there is a character element that is narrower than the width, it is detected and used as a sub-character element as it is.If there is a character element that is wider than the single-byte character width, it is detected and multiple sub-characters with individual widths equal to or less than the single-byte character width are detected. Recognizes a character element dividing unit that divides into character elements and the sub-character elements generated by the character element dividing unit as if they constitute a single character with the character elements or sub-character elements of the preceding and following character strings. A character element image that is cut out by connecting so that the target width after connection is as wide as possible within the range that does not exceed the half-width character width (if the recognition direction is vertical, the concept includes the length in the vertical direction) A half-width character candidate image, a character candidate generation unit that cuts out a character element image that is cut out by connecting so that the width after connection is as wide as possible within a range that does not exceed the full-width character width, and The half-width character candidate image and the full-width character candidate image generated by the character candidate generation unit are temporarily recognized as a single character, and a recognition unit that calculates an evaluation value indicating the corresponding character code and accuracy, and the recognition unit. Substitution of the evaluation values obtained for the half-width character candidate image and the full-width character candidate image into a predetermined function for comparison, and the character candidate image with the higher evaluation value is judged as the correct character Recognition result evaluation It is characterized by having a section.

According to the second aspect of the present invention, the character width calculation unit is the position or size (including, both) of the character element.
Character print pitch is calculated from the character print pitch, and the character print pitch calculation means for calculating the width of the full-width character and the width of the half-width character from the calculation result is provided. In the invention of claim 3, the character width calculation unit determines the height of the character string image (the size or length of the character string image in the direction perpendicular to the character string,
If the text is written vertically, the width is so-called. ) From the temporary character width calculation means for calculating a temporary character width from the character element extraction unit, and a character element whose error between the width of the character element and the temporary character width is smaller than a predetermined value among the character elements extracted by the character element extraction unit. A continuous portion detecting means for detecting a continuous portion and a character printing pitch are calculated from the distance between the midpoints of the detected continuous character elements in the character string direction, and the width of the full-width character and the width of the half-width character are calculated from the calculation result. And a calculation means for calculating

According to another aspect of the present invention, the character width calculation unit has a temporary character width calculation means for calculating the temporary character width from the height of the character string image and a character element having a width smaller than the temporary character width. In this case, the character elements are tentatively connected within a range in which the width does not exceed the tentative character width, and the error between the tentatively connected character element and the original character element width and the tentative character width is determined, and this is the predetermined value. A continuous portion detecting means for detecting a portion in which smaller character elements are continuous, and a character printing pitch is calculated from the distance between the midpoints of the character elements in the character string direction of the continuous portion detected by the continuous portion detecting means. , And a calculation means for calculating the width of the full-width character and the width of the half-width character from the calculation result.

According to a fifth aspect of the present invention, the recognition unit is a half-width character recognition dictionary used for recognizing the half-width character candidate image generated by the character candidate generation unit, and is generated by the character candidate generation unit. And a full-width character recognition dictionary used for recognizing full-width character candidate images. In the invention of claim 6, the recognition unit determines whether or not the recognition result of the full-width character candidate image is a half-width character divided into a half-width character and a half-width character. It has a half-character judgment means and a priority evaluation means for giving priority to the evaluation value of the full-width character candidate image when the recognition result evaluation section evaluates if such a character is judged by the half-character judgment means. It is characterized by that.

In the invention of claim 7, when the character recognition device determines whether the character type is a full-width character or a half-width character, it is temporarily determined to be the same type as the character that is the recognition target immediately before that. When character recognition is performed and if it is determined that the corresponding character does not exist (including, if the evaluation value is very low), it is determined that it is the opposite character type and each part is controlled to perform character recognition. It is characterized by having a full-width half-width tentative judgment control unit.

According to the invention of claim 8, when the character recognition device determines whether the character type is a full-width character or a half-width character, the characters within a predetermined number of characters before and after the character are already recognized as full-width or half-width characters. If it exists, it is characterized by having a full-width half-width determination control unit that controls each unit so that it is used as a determination material. In the invention of claim 9,
For each character image determined to be a single character by the recognition result evaluation unit, a character-to-character space is determined based on whether each character is a full-width character or a half-width character and the midpoint interval between two consecutive characters in the character string direction. And a margin adding section that inserts a margin code at a corresponding position in the output character code based on the detection result of the margin detecting section.

[0014]

With the above construction, in the invention of claim 1, the character string image is cut out from the input document image, and then the individual character images are further cut out from each character string image, and the individual character images are recognized on this. Then, in the character recognition device for converting to the corresponding character code, the following operations are performed.

The character element extraction unit extracts a character element which is a block of pixels forming a character from the character string image. The character width calculation unit calculates the width of the full-width character and the width of the half-width character by a predetermined means from the position or size of the character element extracted by the character element extraction unit. If the character element division unit has a character element narrower than the half-width character width calculated by the character width calculation unit among the character elements extracted by the character element extraction unit, the normal projection shadow of the pixel in the character string direction is not detected. It is detected by the presence or absence of a continuous part, etc., and it is used as it is as a sub-character element, and if there is a character element wider than the half-width character width, it is detected and divided into multiple sub-character elements whose individual width is less than the half-width character width. To do. The character candidate generator cuts out the sub-character elements generated by the character-element division unit by connecting character elements that connect the sub-character elements as much as possible within the range that the width after connection does not exceed the half-width character width. The image to be cut is defined as a half-width character candidate image, and an image cut out by a connected character element formed by connecting sub-character elements is cut out as a full-width character candidate image as long as the width after connection does not exceed the full-width character width. The recognition unit temporarily recognizes the half-width character candidate image and the full-width character candidate image generated by the character candidate generation unit as a single character, and then evaluates the corresponding character code and accuracy by pattern matching or the like. Calculate the value. The recognition result evaluation unit substitutes the evaluation values obtained for the half-width character candidate image and the full-width character candidate image in the recognition unit into a predetermined function and compares them, and the character candidate image with the higher evaluation value of the result is selected. Judge as the correct character.

According to the second aspect of the invention, the character printing pitch calculating means in the character width calculating section is arranged such that the mutual positional relationship of the character elements or the size (width) in the character string direction and the size in the direction orthogonal thereto ( The character print pitch is calculated from the height) and the relative size to the line spacing, and the width of the full-width character and the width of the half-width character are calculated from the calculation result. In the invention of claim 3, the temporary character width calculating means in the character width calculating unit calculates the temporary character width from the height of the character string image. Similarly, the continuous portion detecting means detects, among the character elements extracted by the character element extracting portion, a portion in which the character elements whose error between the width of the character element and the temporary character width is smaller than a predetermined value are continuous. Similarly, the calculation means calculates the character print pitch from the distance between the midpoints of the detected consecutive character elements in the character string direction, and calculates the width of the full-width character and the width of the half-width character from the calculation result.

In the invention of claim 4, the temporary character width calculation means in the character width calculation unit calculates the temporary character width from the height of the character string image. Similarly, when the character elements having a width smaller than the calculated temporary character width are continuous, the continuous portion detecting means temporarily connects the character elements within a range in which the width does not exceed the temporary character width, and the temporarily connected character elements are connected. Further, the error between the width of the original character element and the width of the temporary character is calculated, and the portion where the character elements whose values are smaller than the predetermined value are continuous is detected. Similarly, the calculation means calculates the character print pitch from the distance between the midpoints of the character elements in the character string direction of the continuous portions detected by the continuous portion detection means, and calculates the full-width character width and half-width character width from the calculation results. To do.

In the invention of claim 5, the half-width character recognition dictionary in the recognition section is used for recognizing the half-width character candidate image generated by the character candidate generation section by pattern recognition or the like. Similarly, the full-width character recognition dictionary is used to recognize the full-width character candidate image generated by the character candidate generation unit. In the invention of claim 6, the half-character determining means in the recognition unit determines whether the half-width character candidate image recognition result is a half-width character divided in the character string direction into a half-width one-character character. to decide. Similarly, when the half-character determining unit determines that the character is such a character, the evaluation value of the full-width character candidate image is prioritized in the evaluation by the recognition result evaluation unit.

According to the invention of claim 7, when it is determined whether the character type is a full-width character or a half-width character, it is assumed that the full-width half-width tentative determination control unit is the same type as the character immediately before the recognition target. If it is judged that the character does not exist, it is judged to be the opposite character type and character recognition is performed if it is judged that the corresponding character does not exist (including, cutoff when evaluation value is low). Control each part to perform.

According to the invention of claim 8, when determining whether the character is a full-width character or a half-width character, the full-width half-width judgment control unit determines that the characters within a predetermined number of characters before and after the full-width character have already been recognized as full-width or half-width characters. If it exists, it is controlled to use it as a judgment material. In the invention of claim 9, in each character image that the margin detection unit has determined to be a single character by the recognition result evaluation unit, two characters that are continuous with whether each character is a full-width character or a half-width character The space between characters is detected from the midpoint spacing in the column direction. Similarly, the margin adding section inserts a margin code at a corresponding position in the output character code based on the detection result of the margin detecting section.

[0021]

DESCRIPTION OF THE PREFERRED EMBODIMENTS A character recognition device according to the present invention will be described below based on embodiments. In the following embodiments, characters such as portrait, width, etc., a normalization unit that corrects large characters such as characters in the heading field but full-size characters, etc. into standard characters, output characters after recognition The output part, and further the structural parts of the entire system such as the voicing part that translates into a foreign language or utters based on the output result, the user performs various operations, and inputs the conditions known in advance. Needless to say, it has an input operation unit for performing the operation. However, since these are not directly related to the gist of the present invention, illustration and description thereof will be omitted. (First Embodiment) FIG. 1 is a block diagram of the first embodiment of the present invention. In the figure, reference numeral 1 is an image input unit to which an image is input. A character string extraction unit 2 detects the position of the character string from the document image read by the image input unit 1 and cuts out the character string image. Reference numeral 3 denotes a character element extraction unit that extracts a character element, which is a block of pixels forming a character, from the character string image cut out by the character string extraction unit 2. A character width calculation unit 4 calculates the widths of full-width characters and half-width characters from the positions or sizes of the character elements extracted by the character element extraction unit 3. In the character elements 5 extracted by the character element extraction unit 3, a character element having a width narrower than the half-width character width calculated by the character width calculation unit 4 is detected as a sub-character element and the half-width character A character element dividing unit that detects a character element wider than the width and divides the character element into a plurality of sub-character elements each having a width equal to or smaller than the half-width character width. 6 is a half-width character image for the sub-character element generated by the character element division unit 5 and is an image cut out by the connected character element in which the sub-character elements are connected as much as possible within a range in which the width after connection does not exceed the half-width character width. A character candidate generation unit that cuts out, as a candidate image, an image cut out by a connected character element in which sub-character elements are connected as much as possible within a range in which the width after connection does not exceed the full-width character width as a full-width character candidate image. Reference numeral 7 denotes a recognition unit that temporarily recognizes the half-width character candidate image and the full-width character candidate image generated by the character candidate generation unit 6 as candidate characters, and calculates the corresponding character code and the evaluation value that is the probability of correctness of recognition. . 8 is a recognition result evaluation for comparing the evaluation values obtained for the half-width character candidate image and the full-width character candidate image by the recognition unit based on a predetermined function, and determining the character candidate image with the higher comparison result as a correct character. It is a department. 9 is correct 1 in the recognition result evaluation unit 8
For each character image that is determined to be a character, whether each character is a full-width character or a half-width character, and a space existing between the characters is detected from the midpoint interval between two consecutive characters in the character string direction. This is a margin adding section that inserts a margin code at a corresponding position in the output character code. Reference numeral 10 is a recognition result output unit that outputs a recognition result.

The operation of the character recognition apparatus configured as described above will be described below by taking a horizontal input image as shown in FIG. 2 as an example. In the figure, the lower two circles represent some characters that are not used directly in the explanation of the operation of this embodiment. The image input from the image input unit 1 has 1 pixel forming a character and 0 background pixel other than the character.
Is stored in an image memory (not shown) as binary data. In the character string extraction unit 2, first, the normal projection shadows of the pixels forming the characters in the vertical and horizontal directions are obtained from the document image stored in the image memory, and the normal projection shadows of the pixels forming the characters in both directions are calculated. The width of the regular projection image and the interval between the regular projection images are compared to determine whether the input document image is a vertical writing document or a horizontal writing document (for this technique, for example, Japanese Patent Application No. 60-7763).
No. 3). Then, the document image is divided into blocks parallel to the character string direction, the character part projection in the character string direction is obtained for each block, and the rectangular area sandwiched between the block boundary coordinates and the projection start and end coordinates is divided into that block. By connecting the character strings of each block and obtaining the coordinates of the character strings, by checking for duplication in the direction perpendicular to the character strings of the character strings of each block,
Cut out a character string image. This is shown in FIG. 3A (this technique is disclosed, for example, in Japanese Patent Application No. 60-106404, "Character Recognition Device"). The character element extraction unit 3 extracts a character element from the character string image FIG. 3 (a) cut out by the character string extraction unit 2. 3B, the character string of the character string image is projected in the vertical direction, and a continuous portion of the normal projection shadow of the character portion is set as the character string element. Find the rectangular coordinates. this,
It is shown as s1, s2, ..., S10 in FIG. The character width calculation unit 4 calculates the widths of full-width characters and half-width characters from the coordinates of the character elements extracted by the character element extraction unit 3. There are several possible methods for calculating the character width in the character width calculation unit 4, and three methods adopted in this embodiment will be described below.

The first method will be described with reference to FIG. The height H of the character string is multiplied by a predetermined constant α, and the temporary character width W
Calculate *. Generally, in a printed Japanese document, the aspect ratio of many characters is almost 1, so α takes a value close to approximately 1. However, it goes without saying that there is a slight difference depending on the content of the normalization process or the correction performed according to the printed typeface. Next, there are vertically written "1" (the beginning of the kanji and the kanji number 1) and horizontally written "1 (the kanji character and the Arabic number 1)", so the character width W * is almost the same width. Elements (s3, s5, s9, s10 in the figure)
Width W of the widest character element (s3 in the figure)
The full-width character width Wz is obtained by multiplying max by a predetermined constant β. That is, the width of each character element is Wi (i = 1, 2, ...)
Then, the full-width character width Wz is expressed by the following equation.

Wmax = max {Wi | Wi / W * ˜1} Wz = β1 · Wmax (where β1˜1) where max {U} represents the maximum value of the set U. In addition, from A to B indicates that A and B have almost the same value. The half-width character width Wh is calculated by multiplying the full-width character width Wz by a predetermined constant γ.

Wh = γ · Wz (where
γ to 0.5), the second method will be described with reference to FIG. In the same manner as the first method, the temporary character width W * is calculated, and among the character elements, a plurality of character elements having a width substantially equal to the temporary character width W * are present close to each other (in the figure, The average value or the maximum value of the distances (d1 in the figure) between the midpoints in the character string direction of s9, s10) is set as the character printing pitch P, and the full-width character width Wz and the half-width character width Wh are calculated by the following formulas, respectively. . This is because the longest possible number of characters are evaluated and the change of the character print pitch is not flexibly dealt with.

Wz = β2 · P (where
β2-1 and β2 <1) Wh = γ · Wz (where γ˜0.
5). The third method will be described with reference to FIG. Similar to the first method, the temporary character width W * is calculated, and the temporary character width W * is calculated.
When character elements with smaller width are continuous (in the figure, s6
In s7), character elements are tentatively connected within the range where the width does not exceed the temporary character width W * (s' in the figure) and are treated as a single character, so to speak, and the character width is approximately equal to W *. Average of the distances between the midpoints (d1, d2 in the figure) in the character string direction of elements (s6, s' and s9, s10 in the figure) where multiple elements or tentatively connected character elements exist in close proximity The value or the maximum value is set as the character print pitch P, and the full-width character width Wz and the half-width character width Wh are calculated by the following equations. This is also for the purpose of evaluating characters that are as long and large as possible.

Wz = β2 · P (where
β2 to 1 and β2 <1) Wh = γ · Wz (here, γ to 0.5) Which of the above three methods is to be used depends on the mode of the original document to be processed, that is, Whether or not the character spacing is relatively close, such as newspaper, or whether or not the character spacing is relatively large as in the specification of the patent application,
Needless to say, this is done depending on whether or not full-width Kanji and half-width numbers or alphabets are used together as in an academic paper, and whether or not they are unequal intervals. Further, it goes without saying that the values of the constants such as γ and β2 are appropriately selected according to the document to be recognized.

The operation performed by the character element dividing unit 5 for each character element si of the character elements extracted by the character element extracting unit 3 will be described with reference to FIG. The half-width character width calculated by the character width calculation unit 4 is represented by Wh, and the character element s
Let i be the width of i. First, the division number N is determined. Number of divisions N = [Wi / Wh + 0.5]. here,
[X] is a Gaussian symbol and represents the maximum integer value that does not exceed X.

Next, s4, s5 in FIG.
When the number of divisions N> 1 as in s7 and s10, the character element si of interest as a recognition target is divided into N pieces with a uniform width. 5 (b), s1, s2, s3, s6, s
When the number of divisions N = 1, like 8 and s9, nothing is done.

Next, referring to FIG. 6, with respect to the sub-character elements generated by the character element dividing section 5, the character candidate generating section 6, the recognizing section 7, and the recognition result in order from the sub-character element at the beginning of the character string. The contents of the processing of performing the processing in the evaluation unit 8 and determining the recognition result will be described. First, in the character candidate generation unit 6, paying attention to the sub-character element s1 at the beginning of the character string, the half-width character that is cut out by the connected character element in which the sub-character elements are connected as much as possible within a range in which the width after connection does not exceed the half-width character width Wh The character candidate image c11 and the full-width character candidate image c21 that is cut out by the connected character elements connecting the sub-character elements as much as possible within the range in which the width after connection does not exceed the full-width character width Wz are generated. Then, the recognition unit 7 recognizes the character after making both character candidate images into a single character image, and calculates the character code and the evaluation value. (Note that this evaluation technique is disclosed, for example, in Japanese Patent Application No. 63-312288, “Character recognition method”.) At that time, the half-width character candidate image is a half-width character for improving the recognition speed and accuracy. The recognition is performed using the recognition dictionary, and the full-width character candidate image is recognized using the full-width character recognition dictionary. In the figure, the character code is “8 (Arabic numerals)” for the half-width character candidate image c11, and the evaluation value = 37.
For the full-width character candidate image c21, the character code is "Aki" and the evaluation value is = 61. The recognition result evaluation unit 8 compares these results, and determines that the character candidate image with the higher evaluation is a character.

This comparison may be performed simply by the size of the evaluation value of the recognition result (recognition evaluation value), but in the present embodiment, in order to obtain a more accurate result, the evaluation value is changed depending on the condition. The following function operation is performed, and the final comparison is performed based on the magnitude of the evaluation value (final evaluation value) obtained as a result. Now,
The recognition evaluation value of the half-width character candidate image is represented by v1, the recognition evaluation value of the full-width character candidate image is represented by v2, the final evaluation value of the half-width character candidate image is represented by V1, and the final evaluation value of the full-width character candidate image is represented by V2.

First, the recognition result of the full-width character candidate image is
The first half of one character, for example, horizontal writing "ka" and "eighth" have the same shape as the left half "i" and "no", respectively. If the character has the same shape as one half-width character, a value obtained by multiplying the recognition evaluation value v2 of the full-width character candidate image by a constant δ1 (here, δ1> 1.0) is the final value of the full-width character candidate image. Evaluation value V2
And

That is, V2 = δ1 × v2.
Next, in other cases, a value obtained by multiplying the recognition evaluation value v2 of the full-width character candidate image by a constant δ2 (here, 1.0 <δ2 <δ1) is set as the final evaluation value V2 of. That is, V2 = δ2 × v23.

In the half-width character candidate image, the recognition evaluation value v1 is used as it is as the final evaluation value V1. That is, V1 = v
14 If V1 <V2, the full-width character candidate image is determined to be a character. If V1 ≧ V2, the half-width character candidate image is determined to be a character. This is because weighting is performed so that pixels that are as long as possible are evaluated and recognized.

In the case of FIG. 6, the final evaluation value V1 of the half-width character candidate image c11 is 37, the final evaluation value V2 of the full-width character candidate image c21 is 67, and V1 <V2. And the character code “Aki” is the recognition result. After the recognition result of one character is confirmed as described above, the same process is repeated for the sub-character elements that are successively adjacent to each other. As a recognition result, "Showa 35" displayed in a bold frame in the character code column in the figure is can get. This state is shown in FIG. The character candidate generation unit 6
In the case where the half-width character candidate image c11 and the full-width character candidate image c21 match each other, the full-width character is obtained without performing other recognition processing.

When the recognition result is confirmed for all the characters in the character string, the margin adding section 9 determines whether or not there is a margin.
When the blank space is detected, the blank space is inserted at the corresponding position in the output character code. This process will be described with reference to FIG. In this figure, (a) is (a) of FIG. 3 and FIG.
(B) is added for explanation. It is represented as each character candidate image ci (i = 1, 2, ...) Characterized by the recognition result evaluation unit 8, and double-byte characters are marked with *. The margin adding unit 9 sequentially calculates, for each of the character candidate images ci that have been recognized as characters by the recognition result evaluation unit 8, a midpoint distance di between the next character candidate image ci + 1 and the character string in the character string direction. The presence / absence of a margin is determined as follows. Where P is the character print pitch, and ε1, ε
2 and ε3 are constants of about 1.0, 0.75, and 0.5, respectively. However, as the specific value, another value is adopted according to the print content.

1. When both characters ci and ci + 1 are full-width characters, if di> ε1 · P, there is a space between the characters ci and ci + 1. 2. When one of the characters ci and ci + 1 is a full-width character and the other is a half-width character, if di> ε2 · P, the characters ci and ci +
There is a blank space between areas 1. 3. When both characters ci and ci + 1 are half-width characters, di> ε
If it is 3.P, there is a blank space between the characters ci and ci + 1.

4. In cases other than the above, there is no margin between ci and ci + 1. In (a) of FIG. 7, since the distance d5> ε2 × P between the middle points of the full-width character c5 * and the half-width character c6, a blank space code is inserted between the fifth and sixth characters of the recognition result. On the other hand, in FIG. 7B, the width of the margin between the characters c1 * and c2 * is wide, but since both characters are full-width characters, the distance d1 <ε1 between the midpoints.
XP, and it is determined that the margin is not present. (Second Embodiment) The basic construction of the second embodiment of the present invention is the same as that of the first embodiment. Therefore, in the present embodiment, description of the configuration, action, etc. of each part will be omitted based on the overall configuration diagram, and the principle, purpose, configuration, action, effect of only the unique part will be described. About the principle.

In a normally printed document, it is common to use only one full-width or one half-width character.
Further, even if both types of characters are used together, one of full-width and half-width is predominant, and the other type of characters is used exceptionally. In addition, when the other type of character is used exceptionally, this exceptional character is used continuously in most cases. Therefore, when deciding whether this is a full-width or half-width for a certain character, if there are already determined characters before and after that, this can be used for the determination. About the purpose.

Not only is the character recognition speed improved and the half-width character recognition dictionary and the full-width character recognition dictionary are used more effectively, but the portions printed with different types of characters are:
The performance of the system as a whole is improved by subjecting it to translation or the like separately from other parts or independently. About the configuration.

At the beginning or at the beginning of the sentence, the character type is judged by the constituent elements (requirements) of the first embodiment according to the inventions of claims 1 to 6, and after the judgment of several character types is confirmed. Tentatively determines that the following character is of the same type as the character that has already been determined, and performs character recognition using only one of the full-width character recognition dictionary and the half-width character recognition dictionary. If unrecognizable characters appear, character recognition is performed using the other dictionary. If it is still unrecognizable, the characters are determined by the same constituent elements (requirements) as in the first embodiment. together,
This is called to the user's attention. About the effect.

This is an inside-out of the purpose, and not only the type of the character is judged and the character is recognized, but the performance of the entire system is improved. The present invention has been described above based on the embodiments,
Needless to say, the present invention is not limited to the above embodiment. That is, the following is also included in the present invention.

(1) For convenience of manufacturing, one constituent element (requirement, step) described in the claims is plural. On the contrary, a plurality of items are set to 1. Alternatively, these are appropriately combined. (2) The image cutting-out means has a built-in storage unit such as a floppy disk that can be taken out and mounted, and is separate from other means.

The other storage means can take out the storage part and mount it. Similarly, the output section outputs to a floppy disk or the like. As a result, the portability is improved, the expensive character recognition unit body and the printing unit are effectively used, and they are used together with other devices. (3) Characters to be recognized are not limited to Kanji and Arabic numerals, and they may be Kana (Kana), Hangul characters, alphabets, or a combination of these and Kanji, or a symbol such as "=". Of course, it also includes various symbols.

(4) The writing order of characters is not limited to any one of left horizontal writing, right horizontal writing, and vertical direction, and the user also has a function to input this in advance. Further, in the case of the vertical direction, there are two types from left to right and right to left, and this is also the same.

(5) The constants β1, β2, ζ1, ζ2, γ, etc., are the types of documents such as newspapers, collections of papers, specifications of patent applications, etc., by the user who visually recognizes the characters and contents of the processed document. It is possible to input in advance by making a rough guess based on
In addition, a function is added that allows the user to change or correct the temporary recognition result. (6) Similarly, the underline or upper and lower lines are marked with a predetermined marker on the corresponding part of the character on the sheet before being read, and the position of half-width characters or special symbols is input on the device side. The function that enables it is added. This may be useful when applied to the reading system of newspapers for the visually impaired and school children, and the translation system of foreign language articles in fields where special symbols are often used.

(7) The translation system or the like detects that the whole system cannot translate well, or the user knows that there are many errors in the character recognition results. A function for correcting and learning the recognition process is added by changing various constants automatically or by an instruction. (8) The evaluation method or the recognition means itself for character recognition is not limited to pattern recognition with a dictionary, and other methods such as the decision tree method (Japanese Patent Application No. 5-68586 “Decision tree type character recognition device”) may be used. It is adopted.

(9) When one of the divided character elements in the full-width character is determined not to form a half-width character from the result of collation as a half-width character or before that, from the normal projection shadow length (for example, FIG. A) (S1, S2)) in c) is provided with a function for determining a full-width character. vice versa,
When it is determined that the half-width character is not a half-width character based on the result of collation as a half-width character or before the orthographic projection length (for example, the day (S10) in FIG. 3C), it is determined to be a full-width character. Has been added.

When these functions are combined with the functions of claim 7 or claim 8, the effect is increased. (10) In European and American documents, a blank space is provided between each word, and the blank space is adjusted so that characters are placed at both ends of a line. Therefore, with respect to the blank space between words instead of between characters, a function capable of coping with this may be added.

[0050]

As described above, according to the present invention, a character element dividing unit that temporarily divides an extracted character element into half-width characters is connected to the divided character element, and a half-width character candidate image and a full-width character candidate image are combined. A character candidate generation unit for cutting out and a recognition result evaluation unit for recognizing each of the half-width character candidate image and the full-width character candidate image and selecting the one having a higher evaluation value as a correct character are provided. Therefore, for each character, recognition is performed at most twice, that is, recognition of a half-width character candidate image and recognition of a full-width character candidate image, so that even a document printed in a mixture of full-width and half-width or a document with an irregular pitch can be recognized. It is possible to cut out characters accurately. Furthermore, the characters are cut out while paying attention to the fact that English characters in Japanese, German sentences and vice versa, and double-byte characters used for documents consisting of single-byte characters to call attention are often used consecutively. It becomes possible. As a result, it becomes possible to flexibly deal with the change of the character type and the like while recognizing long and many characters. Therefore,
Accurate character recognition is performed. Further, since it is possible to determine whether the character is a full-width character or a half-width character, it is possible to accurately detect the margin.

[Brief description of drawings]

FIG. 1 is a configuration diagram of a character recognition device according to an embodiment of the present invention.

FIG. 2 is an example of an input document image.

FIG. 3 is a diagram for explaining a character element extraction process.

FIG. 4 is a diagram for explaining a character width calculation process.

FIG. 5 is a diagram for explaining character element division processing.

FIG. 6 is a diagram for explaining candidate character image generation processing, recognition processing, and recognition result evaluation processing.

FIG. 7 is a diagram illustrating a margin adding process.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 image input section 2 character string extraction section 3 character element extraction section 4 character width calculation section 5 character element division section 6 candidate character image generation section 7 recognition section 8 recognition result evaluation section 9 margin addition section 10 recognition result output section

Continuation of the front page (72) Inventor Satoshi Emura 1006 Kadoma, Kadoma City, Osaka Prefecture Matsushita Electric Industrial Co., Ltd.

Claims (9)

[Claims] 1. A character string image is cut out from an input document image, and then individual character images are further cut out from each character string image, on which individual character images are recognized and converted into corresponding character codes. In the character recognition device, a character element extraction unit that extracts a character element that is a group of pixels forming a character from the character string image, and a full-width character from the position or size of the character element extracted by the character element extraction unit. A character width calculation unit that calculates the width and the width of half-width characters by a predetermined means, and a character element that is narrower than the half-width character width calculated by the character width calculation unit among the character elements extracted by the character element extraction unit. If there is a character element with a width wider than the half-width character and it is detected as it is, it is detected as it is and the character element is divided into multiple sub-character elements whose individual width is equal to or less than the half-width character width. For the split part and the sub-character elements generated by the character element splitting part,
Make the width as wide as possible within the range not exceeding the above-mentioned half-width character width, assuming that the character elements or sub-character elements of the preceding and following character strings are regarded as constituting a single character and are connected and recognized. The character element image that is cut out by connection is set as a half-width character candidate image, and the character element image that is cut out by connection is set as wide as possible within a range that does not exceed the full-width character width as a full-width character candidate image. The character candidate generation unit to be cut out, and the half-width character candidate image and the full-width character candidate image generated by the character candidate generation unit are temporarily recognized as a single character, and a corresponding character code and an evaluation value indicating accuracy are calculated. A recognition unit, and the evaluation value obtained for the half-width character candidate image and the full-width character candidate image in the recognition unit is substituted into a predetermined function and compared,
A character recognition device, comprising: a recognition result evaluation unit that determines a character candidate image having a higher evaluation value as a result to be a correct character. 2. The character width calculation unit has a character print pitch calculation means for calculating a character print pitch from the position or size of a character element and calculating the full-width character width and the half-width character width from the calculation result. The character recognition device according to claim 1, wherein: 3. The character width calculation unit calculates a temporary character width from the height of a character string image, and a width of a character element among the character elements extracted by the character element extraction unit. And the temporary character width have an error smaller than a predetermined value. A continuous portion detecting means for detecting a continuous portion of character elements, and a character printing pitch from the distance between the midpoints in the character string direction of the detected consecutive character elements. 2. The character recognition device according to claim 1, further comprising: a calculation unit that calculates the width of the full-width character and the width of the half-width character from the calculation result. 4. The character width calculation unit calculates a temporary character width from the height of a character string image, and a temporary character width calculation means, and when a character element having a width smaller than the calculated temporary character width is continuous. , The character elements are temporarily connected within a range that does not exceed the temporary character width, the error between the width of the temporarily connected character element and the original character element and the temporary character width is calculated, and these values are set to predetermined values. A continuous portion detecting means for detecting a portion where a smaller character element is continuous, and a character printing pitch is calculated from the distance between the midpoints in the character string direction of the character elements of the continuous portion detected by the continuous portion detecting means, The character recognition device according to claim 1, further comprising a calculating unit that calculates a width of the full-width character and a width of the half-width character from the calculation result. 5. The recognition unit includes a half-width character recognition dictionary used for recognizing the half-width character candidate image generated by the character candidate generation unit, and a full-width character candidate image generated by the character candidate generation unit. The character recognition device according to claim 1, 2, 3, or 4, further comprising a full-width character recognition dictionary used for recognition. 6. The half-character determination unit determines whether the recognition result of a full-width character candidate image is such that half of a single character divided in the character string direction has the same shape as one half-width character. And a priority evaluation unit that prioritizes the evaluation value of the full-width character candidate image when the recognition result evaluation unit evaluates the character if the half-character determination unit determines that the character is such a character. The character recognition device according to claim 1, claim 2, claim 3, claim 4, or claim 5. 7. The character recognition device, when judging whether the character type is a full-width character or a half-width character, performs a character recognition after tentatively determining that the character type is the same as the character to be recognized immediately before. However, when it is determined that the corresponding character does not exist, it has a full-width half-width tentative determination control unit that controls to perform character recognition by determining that the character type is opposite. The character recognition device according to claim 1, claim 2, claim 3, claim 4, claim 5, or claim 6. 8. When the character recognition device determines whether the type of a character is a full-width character or a half-width character, if a character within a predetermined number of characters before and after that character has already been recognized as a full-width or half-width character, this 2. A full-width half-width judgment control unit for controlling so as to use as a judgment material.
The character recognition device according to claim 2, claim 3, claim 4, claim 5, or claim 6. 9. A midpoint interval between two character strings in the character string direction, which are continuous with respect to whether each character is a full-width character or a half-width character for each character image determined by the recognition result evaluation unit to be a single character. A blank space detection unit that detects a space between characters and a blank space addition unit that inserts a blank space code at a corresponding position in the output character code based on a detection result of the blank space detection unit. The character recognition device according to claim 1, claim 2, claim 3, claim 4, claim 5, claim 6, claim 7, or claim 8, which is characterized.