JPH045779A - Character recognizing device - Google Patents

Abstract

PURPOSE:To offer a character recognizing device with good accuracy and fast processing speed by extracting the feature of the borderline of a recognizing processing area assigned by a operator and changing the recognizing processing area automatically when the feature does not satisfy fixed conditions. CONSTITUTION:A character recognizing device 20 is provided with a picture storing part 30 of picture data S of characters and figures on a medium, a display part 40 of the picture data S, an input part 50 to assign a recognizing processing area among the picture data S, a recognizing part 60 to recognize the characters with cutting out the picture data at a character unit from the assigned recognizing processing area, a boundary feature extracting part 70 to extract the feature of the borderline of the assigned recognizing processing area, an area changing part 80 to change the assigned processing area based on the feature of the borderline extracted at the extracting part 70 and a control part 90 to control respective constituting components. Thus, recognizing processing is executed to the proper recognizing processing area.

Description

[Detailed Description of the Invention] (Field of Industrial Application) The present invention is applicable to character recognition devices, and particularly to character recognition devices of the type in which an operator instructs the character recognition device to perform recognition processing. It's something to hit.

(Prior Art) In some types of character recognition devices, an operator or character recognition device is required to perform recognition processing on an area (hereinafter referred to as -ti).
¥! It is sometimes abbreviated as recognition processing area. ), the recognition process for this area is executed. Therefore, this type of character recognition device includes an image storage section that stores image data of characters and figures on a medium, a display section that displays the image data, an input section that specifies the recognition processing area 11i of the image data, and an input section that specifies the recognition processing area 11i of the image data. The apparatus included a recognition unit that cuts out image data character by character from the specified recognition processing area and recognizes the characters.

Specifically, the input section consists of a pointing device such as a light ben or a mouse, and the display section consists of a CRTi.
It was composed of The recognition processing area is designated by the operator or a pointing device to designate the desired number of layers of the image take displayed on the CRT.

(Problem to be solved by the invention) However, when trying to use a pointing device to indicate a blank space between character lines or between characters in order to specify a recognition processing area, Printed documents have narrow line spacing and character spacing, so it is difficult to give accurate instructions.

Therefore, the recognition processing area often ends up missing part of the character, or conversely, includes an extra area, which can reduce the recognition accuracy of the character recognition device and reduce the recognition processing speed. There was a problem in that it caused a decrease in

FIGS. 4(A) and 4(B) are diagrams showing a variant example thereof.

FIG. 4(A) is a diagram showing an example in which the lower side of the boundary line 11 defining the recognition processing area crosses over the character originally desired to be recognized, resulting in part of the character being missing. If recognition processing is performed on such a recognition target area, the character recognition device will output a recognition result that is misread or rejected. Therefore, the recognition accuracy of the character recognition device is reduced.

Further, Figure 4 (CB) is a diagram showing an example in which the boundary line defining the recognition processing area extends to areas 11i13 that are not the recognition target. When recognition processing is performed on such a recognition processing area, the character recognition device outputs recognition results that include unnecessary results, so it is necessary to correct the recognition results. Therefore, the processing speed of the character recognition device is reduced.

This invention was made in view of these points,
SUMMARY OF THE INVENTION Accordingly, an object of the present invention is to provide a character recognition device t% that can determine whether or not a recognition processing area specified by an operator is appropriate, and if it is inappropriate, change the recognition processing area to an appropriate one.

(Means for Solving the Problem) In order to achieve this object, the present invention provides an image storage section that stores image data of characters and figures on a medium, a display section that displays the above-mentioned image take, and the above-mentioned A character recognition device comprising: an input unit for specifying a recognition processing area of image data; and a recognition unit for cutting out image data character by character from the specified recognition processing area and recognizing characters;
a boundary feature extraction unit that extracts the boundary line features of the recognition processing area specified by the input unit; and a boundary line feature extraction unit that extracts the boundary line features of the recognition processing area specified by the input unit. and an area changing unit that changes the area based on.

In carrying out the present invention, the above-mentioned display section is configured to display the above-mentioned image take, the above-mentioned recognition processing area specified by the above-mentioned input section, and the above-mentioned changed recognition processing area changed by the above-mentioned area changing section. It is preferable to display them in a format that allows them to be visually distinguished, for example, in different colors or brightness, and to display them overlapping each other.

(Function) According to this configuration, even if the operator mistakenly misses a part of the character to be recognized when specifying the recognition processing area, this character recognition device will still be able to recognize the character that should be recognized. Even if you change the recognition processing area to include a character, or conversely specify that it includes an extra part (for example, a part of the character in the adjacent recognition processing area), this character recognition device will not work. Change the recognition processing area to remove unnecessary parts.

(Embodiments) Hereinafter, embodiments of the character recognition device of the present invention will be described with reference to the drawings.

FIG. 1 is a block diagram schematically showing the configuration of a character recognition device 20 according to an embodiment.

The character recognition device 20 of this embodiment includes an image storage unit 30 that stores image data S of characters and figures on a medium, and an image data 87F! , a display section 40 for displaying, an input section 50 for specifying a recognition processing area of the image data S, and a recognition section 60 for cutting out image data character by character from the specified recognition processing area and recognizing characters. , vfy! of the boundary line of the recognition processing area specified by the input unit 50! a boundary feature extraction unit 70 that extracts the recognition processing area specified by the input unit 50 based on the boundary line feature extracted by the boundary line feature extraction unit 70; .about.80.

Here, in this embodiment, the image take S is input to the image storage unit 30 from a photoelectric conversion unit (not shown), and is expressed by using black pixels for the character line portion on the medium and white pixels for the background portion. This is what I did.

Further, the image storage section 30 of this embodiment is constituted by a memory capable of storing an image take in a format that can reproduce the two-dimensional coordinates of characters and figures on a medium.

Further, the display unit 40 of this embodiment is configured to include a display device such as a CRT.

Further, the input section 50 of this embodiment may be, for example, a light ben,
The device is equipped with a pointing device such as a mouse, a keyboard, and the like.

The operator inputs the character recognition device 20 via this input section 50.
Various instructions can be given.

The recognition unit 60 of this embodiment also includes a line number detection unit 62 that cuts out character line data from image data stored in the image storage unit 30, and a character line detection unit 62 that cuts out character line data from the image data stored in the image storage unit 30; The configuration includes a cutting section 64 and an identifying section 66 that compares and identifies the cut out character batanshi data with, for example, a dictionary.

Further, the boundary feature extraction unit 70 of this embodiment includes the input unit 50
Based on the number of black pixels on the boundary line of the recognition processing area specified by , the feature of the boundary line is extracted.

Furthermore, if the characteristics of the boundary line do not satisfy a predetermined condition, the area changing unit 80 of this embodiment sets a plurality of boundary candidate lines in the vicinity of the boundary line, compares the characteristics of these boundary candidate lines, and determines the characteristics. Among the boundary candidate lines that satisfy the above-mentioned predetermined conditions, the boundary line [the closest multiple boundary candidate line is set as the new boundary line.

Further, the control unit 90 controls each component 30 to 80.

Next, details of each of the above-mentioned components 30 to 90 will be explained together with the operation of the character recognition device of the embodiment.

When the image recorder 1w section 30 stores the image data S, the control section 90 reads the image data in the image storage section 10 and transfers this image take and the color data for displaying it on the display section 4o to the display section 40. .

The display unit 40 displays the image data transferred by the control unit 90 on the display in a color specified by the control unit 90.

Note that the input unit 50 can specify any M1 color to be displayed.

Next, in order to specify the recognition processing area of the image data displayed on the display of the display unit 40, the operator uses the pointing device of the input unit 50 or the like to specify the recognition processing area.

In response to this, the input unit 50 outputs the coordinates of the designated recognition processing area to the control unit 90.

In this case, the recognition processing area is specified by the operator via the input section 50 by specifying two points on the display of the display section 40 using a pointing device. Figure 2 shows a concrete example of the results.
It is a diagram schematically showing in a visually visible manner the image data in the image storage section 30]00 and the recognition processing area]02 designated by the input section 50. The rectangular area surrounded by the first point and the second point B is the recognition processing area. However, in this example, due to the operator's sweeping error, the lower part of the boundary line defining the recognition processing area 702 or the top of the character that was originally intended to be recognized was crossed, resulting in part of the character being missing. be.

Next, the control unit 90 outputs to the display unit 40 the coordinates of the recognition processing area input from the input unit 50 and color data for displaying the boundary line of the recognition processing area on the display unit 40.

The display unit 40 displays the boundary line of the recognition processing area on the display according to the coordinates and color indicating the recognition processing area transferred from the control unit 90, overlapping the already displayed image data.

The control unit 90 also causes the boundary feature extraction unit 70 to extract the characteristics of the boundary line that defines the recognition processing area, or in the case of KJ Chiko's embodiment, the characteristics of each side of the four sides that define the recognition processing area. , and outputs the coordinates of each side to the boundary feature extraction unit 70.

The boundary feature extraction unit 70 counts the number of black pixels on each of the four sides input from the control unit 9Q, determines whether the number of black pixels exceeds a predetermined number, and uses the determination result and the coordinates of the four sides. is output to the area changing section 80. Incidentally, the above-mentioned predetermined number is not limited to this, but in this case, it is a number corresponding to 1% of the number of pixels forming the corresponding side. In other words, in the case of the lower side of FIG. 2, the number of pixels is 1% of the number of pixels determined by the X coordinate difference between points A and B.

For such a predetermined number, if it is set on the border line or the background between two character lines, only white pixels will be detected on that border line, so it will be detected unless the predetermined number is exceeded, and the border line will be detected. If the number is set over a character line, black pixels will be detected on the boundary line, so it is also possible that the number exceeds the predetermined number. In this way, it is determined whether the recognition processing area is properly specified.

In the case of the example shown in FIG. 2, the lower side of the four sides constituting the boundary line is separated by a character line, so the number of black pixels exceeds a predetermined number. In this embodiment, the boundary feature extraction unit 7° outputs this information to the area change unit 80 in the form of an area change number. This area change number is used by the area change unit 8 to indicate on which of the four sides the number of black pixels exceeds a predetermined number.
0, if the number of black pixels on the top side of the four sides exceeds a predetermined number, it is set as "1", if it is on the left side, it is set as "2", and if it is on the bottom side, it is set as "1". 4”,
Similarly, the right side is assumed to be "8". Therefore, if the number of black pixels on all four sides exceeds the predetermined number, the area change number will be "]5," while if the number of black pixels does not exceed the predetermined number on all four sides, the area change number will be "0".

If the area change number input from the boundary feature extraction unit 70 is rQJ, the area change unit 80 controls the coordinates of the four sides by controlling the m unit 9.
Output to 0 as is. In addition, if the area change number is "other than 0", the area change unit 80 sets a plurality of boundary candidate lines parallel to the sides according to the area change number, extracts the number of black pixels on each boundary candidate line, and extracts the number of black pixels on each boundary candidate line. The coordinates of a boundary candidate line that is closest to the first boundary line among the boundary candidate lines whose number of pixels does not exceed the above-mentioned predetermined number are output to the control unit 90. Ivy 2
To explain specifically using the example shown in the figure, the Y coordinate of the lower side is
±1, ±2 for the initially set Y coordinate.・・・・・・
A new boundary line 104 (indicated by a broken line in FIG. 2) is determined from among the boundary candidate lines obtained by changing the boundary lines up to .

Therefore, in this example, the recognition processing area is changed to be enlarged in the lower side direction. Incidentally, in this embodiment, extraction of the features of each boundary candidate line may be performed in the area changing section 80, or may be performed in the boundary feature extraction section 70, of course.

Next, the control unit 90 transmits to the display unit 40 the new coordinates indicating the recognition processing area input from the area changing unit 80 and the color data for displaying this new recognition processing area on the display unit 40.
Output to.

The display unit 40 displays the new recognition processing area in the specified color on the display, overlapping the recognition processing area that is already displayed and specified by the operator.

FIG. 3 shows an example of this display, which corresponds to the conditions shown in FIG. 2. In FIG. 3, ]10 is the display screen of the display device, 112 is image data,
114 is the boundary line of the recognition processing area specified by the operator;
116 is a boundary line of a new recognition processing area changed by the character recognition device. In the case of this embodiment, the image data 112 is displayed in black, and the boundary line of the recognition processing area specified by the operator]1
4 is displayed in blue, and the border line of the new recognition processing area] 16 is displayed in red.

Further, the control unit 90 outputs new coordinates indicating the recognition processing area inputted from the stop area changing unit 80 to the line position detection unit 62 in the recognition unit 60.

The line position detection unit 62 moves an area (recognition processing area) of the image storage unit 30 given by the coordinates input from the control unit 90 in a predetermined direction (in the case of this embodiment, a horizontally written document is the object to be recognized). (horizontal direction), counts the cumulative number of black pixels for each line, extracts the peripheral distribution, detects the line position based on this peripheral distribution, and outputs this line position M to the character cutting unit 64. do.

Next, the character cutting unit 64 scans the line inclination area in a predetermined direction (in the case of this embodiment, the vertical direction) based on the line position input from the line position detection unit 62, and blackens each column by column. The cumulative number of pixels is counted to extract the peripheral distribution, character slam data is cut out for each character based on this peripheral distribution, and this character slam data is output to the identification section 66.

Next, the identification unit 66 extracts the characteristics of the character slam data inputted from the character extraction unit 64 using a conventionally used suitable method, and controls the identification result by comparing the characteristics with the word 1 prepared in advance. 90.

Next, the control unit 90 outputs the identification result input from the identification unit 66 to the display unit 40. The display unit 40 displays the identification results on a display. The operator checks the recognition results displayed on the display, and if there is a misread or rejection, the operator
Correct by 0.

As a result of the operator's confirmation, if the identification has been performed correctly or if the correction has been completed, the control unit 90 outputs the identification result to the terminal 92 as a character name output.

Although the embodiments of the character recognition device of the present invention have been described above, the present invention is not limited to the above-described embodiments, and can be modified in various ways as described below.

For example, in the above-described embodiment, the boundary feature extraction unit is configured to extract the number of black pixels on the boundary line as a feature, but the configuration is not limited to this. For example, the boundary feature extraction unit is configured to convert black pixels to white pixels on the boundary line (
Alternatively, the configuration may be such that the number of changes (from a white pixel to a black pixel) is extracted as a feature. In addition, the boundary feature extraction unit may be configured to extract the black pixel density on the boundary line, for example, to extract the number of black pixels per unit length on the boundary line as a feature, or whether the black pixel density is N@ (N is The configuration may be such that the number of consecutive parts (a set number) or more is extracted as a feature.

Further, in the above-described embodiments, the display section is configured to display the border lines, image data, etc. in different colors, or may be configured to display them in different brightness.

(Nine Achievements of the Invention) As is clear from the above description, the character recognition device of the present invention extracts the feature of the boundary line of the recognition processing area specified by the operator, and extracts the feature of the boundary line of the recognition processing area specified by the operator, and extracts the feature of the boundary line of the recognition processing area specified by the operator, and extracts the feature of the boundary line of the recognition processing area specified by the operator, and extracts the feature of the boundary line of the recognition processing area specified by the operator. If it is not satisfied, the recognition processing area is automatically changed.

Therefore, even if the operator makes a mistake in specifying the recognition processing area, causing part of the character to be recognized to be missing, this character recognition device will adjust the recognition processing area to include the character to be recognized. This character recognition device automatically changes the text and vice versa (even if you specify to include extra parts (for example, part of a character in the adjacent recognition processing area), this character recognition device will remove the extra parts. The recognition processing area is automatically changed so that the recognition processing area is automatically changed.Since the recognition processing is performed on the appropriate recognition processing area, it is possible to provide a character recognition device with high accuracy and high processing speed.

In addition, the display unit is configured so that the recognition processing area specified by the operator and the new recognition processing area can be displayed overlappingly by changing the color and brightness so that the operator can distinguish between the recognition processing area and the new recognition processing area. This is particularly suitable because it is easy to check whether the area is the area intended by the operator or not.

[Brief explanation of the drawing]

! Figure T is a block diagram showing the configuration of the character recognition device of the embodiment, Figure 2 is a diagram showing the relationship between the image take in the image storage unit and the recognition processing area specified by the input unit, and Figure 3 is , a diagram showing a display example corresponding to FIG.
This is a diagram to provide. S...Image take, 20...Character recognition device 30...Image storage unit, 40-...Display unit 50...
- Human power department, 60... Recognition unit 62... Linear word detection unit, 64... Character cutting unit 66... Identification unit, 70... Boundary feature drawing unit 80... Area changing unit , 90...control unit 92...output terminal,
100... Image take 102... Recognition processing area specified by the input unit 110... Display screen,
112--Image data 114--Boundary line of the recognition processing area specified by the operator 116--Boundary line of the new recognition processing area that has been changed. Patent applicant: Oki Electric Industry Co., Ltd. 110 Display screen 112, image tick 114, boundary line 116 of the recognition processing area specified after the operation, and display example corresponding to the changed new recognition processing area boundary line in FIG. 1. :7Figure 3Figure 4

Claims (5)

[Claims] (1) An image storage unit that stores image data of characters and figures on a medium, a display unit that displays the image data, an input unit that specifies a recognition processing area of the image data, and the specified recognition processing area A character recognition device comprising a recognition unit that extracts image data character by character from a character and recognizes characters, comprising: a boundary feature extraction unit that extracts characteristics of a boundary line of a recognition processing area specified by an input unit; A character recognition device comprising: an area changing unit that changes a designated recognition processing area based on the boundary line feature extracted by the boundary line feature extraction unit. (2) If the characteristics of the boundary line do not satisfy a predetermined condition, the area changing unit sets a plurality of boundary candidate lines in the vicinity of the boundary line, extracts the characteristics of these boundary candidate lines, 2. The character recognition device according to claim 1, wherein a boundary candidate line closest to said boundary line among boundary candidate lines satisfying a condition is set as a new boundary line. (3) The boundary feature extraction unit is configured to perform the boundary feature extraction section based on one or more of the following elements: the number of black pixels on the boundary line, the number of changes from black pixels to white pixels on the boundary line, and the density of black pixels on the boundary line. The character recognition device according to claim 1, characterized in that the character recognition device is configured to extract features of the boundary line. (4) The display unit is arranged to overlap the image data, the recognition processing area designated by the input unit, and the changed recognition processing area changed by the area changing unit in a format that allows each to be visually distinguished. 2. The character recognition device according to claim 1, wherein the character recognition device is configured to display the character. (5) The display unit is configured to display the image data, the recognition processing area specified by the input unit, and the changed recognition processing area changed by the area changing unit in different colors or brightnesses. Claim 1 characterized in that
Or the character recognition device described in 4.