JPH03233695A - Character array detector - Google Patents

Abstract

PURPOSE:To easily and surely detect a character array by comparing a minimum value with a threshold while successively shifting a noticed point by calculation of the distance and detection of the minimum value with respect to each character position constituting a character group to discriminate a row position. CONSTITUTION:A picture input device 12, a picture memory 13, a matching processing part 14, a microprocessor unit 17, a memory 23, etc., are provided. The minimum value is compared with the threshold while successively shifting the noticed point by calculation of the distance and detection of the minimum value with respect to the position of each of characters constituting the character group, thereby discriminating a row position. Thus, the character array is easily and surely detected though calculated character positions are not uniform.

Description

[Detailed Description of the Invention] <Industrial Application Field> The present invention relates to a character arrangement detection method used to detect the arrangement of characters to be recognized for each line in an image of a group of characters such as printed characters spanning multiple lines. Regarding equipment.

<Conventional technology> Conventionally, character recognition devices detect breaks in character constituent pixels (black pixels) by taking density distributions in two directions orthogonal to each other in an image of a group of characters, and cut out characters at each break to identify each character. A method has been proposed in which, after detecting the position, a template is applied to each character position to recognize the character and detect the character arrangement.

However, with this method, as shown in the image shown in Figure 5,
If noise 3 exists between characters 1.1 or underline 4 exists across multiple characters 1 as shown in the image shown in Figure 6, each density distribution HT
2, which should be a break is not detected as a break, making it difficult to cut out characters.

The same is true when the character string is tilted and each character is rotated, as in the image shown in FIG.
It is still difficult to cut out the characters.

Therefore, the inventor developed a method for recognizing the position and content of each character by scanning a standard pattern of each character over the entire image of a character group and performing template matching.

According to this method, the position and content of each character can be recognized even if there is noise or underlining, or even if each character has an inclination.

<Problem to be solved by the invention> However, in the case of the second method, the obtained character positions are not necessarily arranged in an orderly manner line by line. It is necessary to correct the

This invention allows even if the detected character positions are irregular,
It is an object of the present invention to provide a novel character array detection device that can easily and reliably detect character arrays line by line.

<Means for Solving the Problems> The character arrangement detection device of the second invention includes a storage means for storing detection data of the position of each character constituting a character group, and a plurality of character positions stored in the storage means. Distance calculating means for determining a point of interest and calculating the distance from the point of interest to another character position, and detecting the minimum value among the calculated values obtained by the distance calculating means and calculating a new character position corresponding to the minimum value. The apparatus includes a point-of-interest extracting means for extracting a point of interest, and a determining means for determining the row position of a new point of interest by comparing the minimum value with a predetermined threshold.

<Operation> For the position of each character constituting a character group, the line position is determined by calculating the distance and detecting the minimum value while sequentially shifting the focus point and comparing the minimum value with a threshold value. Even if the character positions are irregular, it is possible to easily and reliably detect the character arrangement.

<Embodiment> FIG. 1 shows the overall configuration of a character recognition device 11 incorporating a character arrangement detection device of the present invention.

The illustrated device 11 is of a type in which a template of each character to be recognized is scanned over the entire input image of a group of characters spanning multiple lines, and the image input device 12. Image memory 1
3. Matching processing section 14. Microprocessor unit (hereinafter referred to as microprocessor unit)

(abbreviated as MPU)17. The MPU 17 includes a memory 23 and the like, and is connected to an operation section 20 such as a keyboard and a host computer 22 via an external interface circuit 21 .

The image human power device 12 takes in the reference image of the template to be registered in the teaching mode and stores it in the image memory 13, and also inputs the image to be recognized in the recognition mode and provides the image data to the matching processing section 14.

The reference image stored in the image memo IJ 13 is
7 and the image of each character whose area has been specified by the operation unit 20 is stored in the pattern memory 1 as a template.
Registered in 8. Note that the enlargement/reduction processing section 19 is for creating templates with different character sizes.

The pattern matching section 14 includes a pattern matching section 14.
a, a matching degree calculating section 15, and a -number position detecting section 16. Image data for one frame of recognition target is inputted to the pattern matching section 14a from the image input device 12, and A template is loaded by the MPtJ 17, and template matching is performed by sequentially scanning the template over the input image.

The matching degree calculation unit 15 is a part that calculates the number of matching pixels between each constituent pixel of the template and each corresponding pixel of the input image at each scanning position of the template as a matching degree. For example, if the character constituent pixels are black pixels and the other background pixels are white pixels, the number of pixels in which each pixel of the template and the corresponding pixel of the human image are both black pixels or white pixels is counted.

The matching position detection unit 16 compares the matching degree calculated by the matching degree calculation unit 15 with a predetermined threshold value given by the MPU 17, and the matching degree exceeding the threshold value is detected by the matching degree calculation unit 1.
5, the position is detected as a candidate position of the character in the template. In this case, when a large number of candidates are detected for one character position, an integration process such as setting the candidate position with the highest degree of matching as the character position is performed.

By applying such processing to the entire image data for each template, all character positions in the input image are obtained, and the position data is stored in the memory 17.

The table below shows the X and Y coordinates of each character position obtained by the above matching process. In the same table, (X 10. yl
o) 1. ” (X Ill+ y+n) is the character position that matched the first template, (XZ.

y2゜)...(xzn, yzn) respectively indicate the character positions that matched the second template (the same applies hereafter). Further, in the same table, the processed flag is a flag that is turned on when the processing for that character position has been completed, and is turned off when the character position has not been processed.

FIG. 2 shows each character position P I+ P z, . . . , Pl on the human image obtained by the matching process. In the case of the illustrated example, character positions P, ~P11゜P, □~P
24. P25-P36. . . corresponds to one line of the character string, and the MPU 17 then detects the character arrangement line by line by executing the procedure shown in FIGS. 3 and 4.

First, in step O (indicated by rsTOJ in the figure) in Figure 3, the origin 0 (0, 0) of the input image is set as the point of interest, and all processed flags for each character position data shown in the table above are turned on. , sets the line feed flag to ON in the flag area of the memo IJ23. This line feed flag is used to indicate whether or not the current process is immediately after a line feed. In the next step 1, the character position (P in the example of FIG. 2) whose processed flag is on and which is closest to the current point of interest is determined. The procedure for this process is shown in FIG. 4 and will be described in detail later.

In the next step 2, it is determined whether the line feed flag is on or not. In this case, the judgment is affirmative, so proceed to step 7, turn off the line feed flag, and then change the character position to a new focus. point, and the processed flag for that character position is turned off (step 4).

After this, the process returns to step 1, and the processed flag is turned on.
, and the character position closest to the point of interest (P2 in the example of FIG. 2) is obtained by executing the procedure of FIG.

Next, in step 2, it is determined whether the line break flag is on, and in this case, the determination is negative, so the process proceeds to step 3, and the distance between the current point of interest and the obtained character position is a predetermined threshold. It is determined whether the value is greater than or equal to the value. If distance #D
If is smaller than the threshold, step 3 is negative;
Determine that the current point of interest is not at the end of the line, that is, at the line break position,
The above character position is set as a new point of interest on the same line, and the processed flag of that character position is turned off. After this, the process returns to step 1 and the same procedure is repeated.

In this iterative procedure, if the judgment in step 3 is affirmative, the process proceeds to step 5, where it is determined that the current point of interest is the end of the line, that is, the line feed position, and the processed flag is turned off and the line feed flag is turned on. At the same time, the origin ○ (0, O) of the human image is set as the point of interest. In the next step 6, it is determined whether or not all character positions have been checked.If negative, the process returns to step 1 and proceeds to the next line, and the above series of steps is repeated until there are no more character positions to check. Execute repeatedly. For each point of interest other than the origin, its position and arrangement order for each row are stored in memory 2.
Of course, the data may be stored in a separate storage area.

Next, to explain the procedure shown in Fig. 4, in order to find the character position closest to the point of interest, Ste Knobs 1' to 3' in the same figure are connected to all character positions whose processed flag is on. This is a procedure for determining the distance, and step 1' determines whether or not the processed flag of the character position to be checked is on. If it is on, step 712' calculates the distance ID between the point of interest and its character position using the following calculation formula weighted in the vertical direction (vertical direction in FIG. 2).

D=(x, −,)”=w2(y; −yJ)2In the above formula, (x+,)'i) is the coordinate of character position P1,
(XJ, yJ) are the coordinates of the position P of the point of interest, and W is the weight (1 or more).

The next step 13' determines whether the above-described distance calculations have been performed for all character positions, and if the determination is "No", the steps 2 and 2' are repeated. Note that if the processed flag in step 1' is orf, step 2' is skipped.

In this way, if the judgment in step 3' is affirmative, in the next step 4', the minimum distance from the calculated distances to each character position is determined, and for each character position corresponding to that distance, step 3 in FIG. A judgment will be made.

Note that although the above embodiment deals with horizontally written character groups, the character arrangement can also be detected for vertically written character groups using the same procedure.

In this case, it goes without saying that a horizontally weighted calculation formula is used in the distance calculation in step 2' of FIG.

<Effects of the Invention> As described above, the present invention sequentially shifts the focus point by calculating the distance and detecting the minimum value for each character position constituting the character group, and compares the minimum value with a threshold value. Since the line positions are determined, the character arrangement can be easily and reliably detected even if the detected character positions are irregular, achieving the remarkable effect of achieving the object of the invention.

[Brief explanation of drawings]

FIG. 1 is a block diagram of a character recognition device in which the present invention is implemented, FIG. 2 is an explanatory diagram showing character positions on an input image, and FIG.
5 and 4 are flowcharts showing the character arrangement detection procedure, and FIGS. 5, 6, and 7 are explanatory diagrams for explaining the problems of the conventional example. 11...Character recognition device 14...Matching processing unit 17...M P IJ 23...
Memory No. 4 Figure i Sugihi Sashi 11th - Q and Nezu Flow Kate Diagram g#! PItlTr, J-lI, q, h company aft 1)
, 1:i? Company binding map ■ Figure 7'J''qM/!r5M, #, z Liu Jr.

Claims (1)

[Scope of Claims] Storage means for storing detection data of the position of each character constituting a character group; A point of interest is determined for a plurality of character positions stored in the storage means, and a method from the point of interest to other character positions is provided. a distance calculation means for calculating the distance between the two points; a point-of-interest extraction means for detecting the minimum value among the calculated values obtained by the distance calculation means and extracting the character position corresponding to the minimum value as a new point of interest; and the minimum value. 1. A character arrangement detecting device comprising: a determining means for determining a line position of a new point of interest by comparing the line position of a new point of interest with a predetermined threshold value.