JP3476936B2 - Character line direction determination device - Google Patents

Description

Description: BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to, for example, an address reading apparatus for reading address information on a postal matter. The present invention relates to a character line direction determining device that determines the direction of a destination character line to be configured. 2. Description of the Related Art For example, as an apparatus for reading and recognizing printed or typed characters, an address reading apparatus in an automatic mail sorting machine is a typical example. Usually, mails with printed or typed postal codes or address lines are often sent in bulk by large users, and it is necessary to process mails of these users in the same format at once. . Therefore, conventionally, when this kind of mail is read by the address reading device, the vertical or horizontal writing of the address character line is distinguished (that is, the character line direction is changed) so that character recognition becomes easy. Is set in advance. However, when a postal matter of various formats is originally processed in a short time, or when a lot of types of postal matter are mixed, the setting of the writing direction (character line direction) by manual operation must be performed. This is a major problem in improving the processing capability and operation of the address reading device. Conventionally, as a first method for determining the direction of a destination character line described on a postal matter, for example, a line inside a round frame of a post-payment stamp printed on a postal matter, or
There is known a method of determining the writing direction of a destination character line by detecting the typeface of a character. That is, the direction of the address character line is determined based on the fact that the direction of the line inside the round frame of the post-pay stamp or the typeface direction of the character matches the writing direction of the address character line. . [0006] Conventionally, as another second technique, there has been known a method in which the vertical dimension and the horizontal dimension of a character block on a mail are measured, and the larger dimension is set as the vertical direction. I have. This is based on the fact that printed or typed characters have a larger vertical dimension (height) than a horizontal dimension (width) except for specific symbols. [0007] However, in the above-mentioned first method, the direction of the line inside the round frame of the post-paid stamp or the separate stamp, or the typeface direction of the character and the writing direction of the address character line are described. Since the direction of the address character line is determined based on the fact that the character string matches, the postal matter that violates this rule makes an incorrect determination of the direction of the address character line. In addition, the direction of the address character line cannot be determined for postal items without post-pay or post-payment stamps. [0008] In the above-mentioned second method, printed or typed characters, except for specific symbols, are:
Since the direction of the address text line is determined based on the fact that the vertical dimension (height) is larger than the horizontal dimension (width), if the type characters (Kanji etc.) violate this rule, Incorrect line direction determination. Further, if the upper and lower ends of the characters are erroneously detected because the characters are close to each other or the space between the radicals constituting the kanji is large, the direction of the address character line may be erroneously determined. SUMMARY OF THE INVENTION It is an object of the present invention to provide a character line direction discriminating apparatus capable of accurately discriminating the direction of a character line on a read object without being restricted by the prior art. According to the present invention, there is provided a character line direction discriminating apparatus for inputting an image including a character line on a reading material having a character line described in a horizontal direction or a vertical direction. input means, the input image by the image input means
On the other hand, using the differential matrix to
Differentiating means for differentiating each,
And the process of taking the absolute value of each pixel value differentiated in the vertical direction.
The same code that matches the sign of the pixel value by performing
Encoding means and a horizontal code whose codes are matched by the same encoding means.
Maximum point is detected from each pixel value in the vertical and vertical directions
Maximum point detection means and the maximum point detection means
OR of selected horizontal and vertical maxima
Binarizing means for binarizing by means of
For each pixel in the binarized image, the pixel
When an isolated value is shown with a surrounding pixel, that pixel is set to a specific value.
Isolated point removing means for removing isolated points by replacing
Of the image from which isolated points have been removed by the isolated point removing means.
Focusing on the effective pixel, if there is an effective pixel in the vicinity,
By performing the process of making the pixels of
The labeling means to be labeled and the labeling means
For the more labeled image, the smallest containing that image
A circumscribing to make a circumscribed rectangle by creating a small circumscribed rectangle
The circumscribed rectangle is formed by the rectifying means and the circumscribed rectangle.
Of the circumscribed rectangle of small or excessive size in the
The small / large circumscribed rectangle removing means to be removed
The remaining circumscribed rectangle as a result of processing by the large circumscribed rectangle removing means
Clustered images horizontally and vertically
And integrating means for integrating by reduction, to the integration means
And comprising counting means for counting more integrated each direction integration count, respectively, and discriminating means for discriminating the direction of the character line of the reading Butsujo based on each integrated number counted by the counting means. According to the present invention, for example, in mail,
In order to determine the direction of the character line based on the characteristics described in the character, regardless of the direction of the line inside the round frame of the post-pay stamp or the separate mark and the typeface of the character,
Even if they do not exist on the mail, the direction of the address character line can be accurately determined. Since the direction of a character line is determined based on the characteristics of the entire character line instead of analyzing each character, the direction of the character line can be determined regardless of the determination result of each character. . Further, for example, in a postal matter, by being able to determine the writing direction of the address character line of the postal matter,
The recognition direction can be specified at the time of character recognition, and it is not necessary to perform character recognition assuming another writing direction. An embodiment of the present invention will be described below with reference to the drawings. FIG. 1 shows a configuration of a mail address reading device to which a character line direction discriminating device according to the present invention is applied. That is, the address reading device includes a light source 101 that illuminates a mail P as a read material conveyed in the direction of the arrow in the figure, an optical system 102 such as a lens, and a mail P.
An image input unit (image input means) 104 for inputting an image including an address character string on the mail P,
A / D for digitizing the output signal of line sensor 103
The conversion unit 105, the image memory 106 for storing the image data output from the A / D conversion unit 105, the processing region specification unit 107 for specifying the processing region, and the processing region specification unit 107 for the image data in the image memory 106. Blocking processing unit 108 for blocking image data in the processed processing area,
A clustering processing unit 109 that clusters (integrates) the data that has been blocked by the blocking processing unit 108 in the horizontal direction and the vertical direction, and the number of clusters in the horizontal and vertical directions that are clustered by the clustering processing unit 109 And a cluster count / comparator 110 and a cluster count / comparator, which compare the counted numbers of clusters in the horizontal and vertical directions and output the integration direction of the smaller number of clusters as the character line direction. 110
The character recognition unit 111 recognizes a character on the mail P based on the character line direction output from. The details of each section will be described below. The postal matter P is assumed to have address text lines constituting address information written in the horizontal or vertical direction in an address area on the front surface thereof, and are conveyed in the direction of the arrow shown in the figure by a conveyance path (not shown). The image on the conveyed mail P is photoelectrically converted by the line sensor 103, and an analog signal (image signal) corresponding to the image is sent to the A / D converter 105. The A / D converter 105 converts the transmitted analog signal into a digital signal, and stores the digital signal in the image memory 106 as image data. Processing area designation unit 107
Are the block processing unit 108 and the clustering processing unit 1
At 09, an image area to be processed is specified. The block processing unit 108 includes the image memory 1
The image in the area designated by the processing area designation unit 107 of the image stored in 06 is binarized, and is divided into blocks by labeling processing. Here, various methods are conceivable for the binarization processing. For example, binarization is simply performed using a predetermined threshold, the binarization threshold is changed according to the value of an image adjacent to the pixel of interest, For example, the image may be differentiated and then binarized, but in this case, any method may be used. Also, various methods can be considered for the labeling process, but here, the labeling method is not particularly defined. The clustering processing unit 109 clusters the block data divided by the blocking processing unit 108 in a specific direction, in this embodiment, horizontally and vertically. The cluster number counting / comparing unit 110 obtains the number of clusters clustered in the horizontal direction and the number of clusters clustered in the vertical direction in the clustering processing unit 109, compares them, and compares the two. Is output as the character line direction. That is, in this embodiment,
In general, address character lines on postal matter P are characterized in that they are written continuously in a certain direction, and advertisement characters tend to be written in the same direction as the address character lines. Character line direction is determined. The character recognizing section 111 outputs mail P in the direction of the character line output from the cluster number counting / comparing section 110.
Recognize the above address characters and output the recognition result. FIG.
7 shows the configuration of the blocking processing unit 108 in detail. In the block processing described below, processing is performed only on the processing area specified by the processing area specifying unit 107. First, the binarization process, here,
A method of performing binarization by the differential processing unit 501, the encoding processing unit 502, and the local maximum point detection processing unit 503 will be described. The differentiation processing section 501 differentiates the image stored in the image memory 106. This differentiation process uses the differentiation matrices shown in FIGS. 3A and 3C,
This is performed in each of the X direction and the Y direction. Specifically, in the image data in the image memory 106, if the processing area specified by the processing area specifying unit 107 is, for example, an area as shown in FIG.
For the direction, the matrix of FIG.
1,602,603,...
In this case, for code 601, the result is 5 × 1 + 3 ×
(−1) = 2, and the differential value of the pixel coordinate position (0, 0) is “2”. This operation is performed at the pixel coordinate positions (4, 0), (4, 1), (4, 2), (4, 3),
Performed for all pixels except (4, 4). Similarly, in the Y direction, the matrix shown in FIG. 3C is calculated using pixel coordinate positions (0, 4), (1, 4), (2, 4), (3, 4).
4) Except for (4, 4), as shown in FIG.
.., 605, 606,... The encoding processing unit 502 includes a differential processing unit 50
A process for taking the absolute value of the pixel value differentiated by 1 is performed. X
The differential values for the direction and the Y direction are respectively shown in FIG.
If it is assumed to be as shown in FIG. 4C, the same encoding process results in the states shown in FIGS. 4B and 4D, respectively. The maximum point detection processing unit 503 extracts the maximum point from the pixel values processed by the same encoding processing unit 502 and binarizes it. For example, if the result of the same encoding process is X direction, Y direction
Assuming that the directions are obtained as shown in FIGS. 5A and 5C, the method of extracting the local maximum point is, for example, a local maximum point having a pixel value “5” or more.
It is determined as shown in FIG. Here, the pixel detected as the maximum point is “1”, and the other pixels are “0”. Then, the logical sum (O) of the values in the X and Y directions is obtained.
By taking R), a result as shown in FIG. The isolated point removal processing section 504 removes the isolated points of the binarized image obtained by the local maximum point detection processing section 503. Specifically, when a matrix as shown in FIG. 6A appears in the image data, a process of replacing the center pixel with “0” is performed as shown in FIG. 6B. The labeling processing section 505 labels the image processed by the isolated point removal processing section 504. More specifically, focusing on the “1” pixel in the image after the isolated point removal illustrated in FIG. 7A, if the “1” pixel exists near 8, the pixel is labeled with the same label. Perform processing. FIG. 7B shows the result of performing the labeling process on the image of FIG. 7A. As shown, in this case, one label 1001 is created. The circumscribing rectangle processing unit 506 creates the smallest rectangle including the image labeled by the labeling processing unit 505. The circumscribed rectangle of the labeled image as shown in FIG. 8A is a rectangle 1101 as shown in FIG. 8B. At the end of the blocking processing section 108, the minute / excess label removing processing section 507 removes a circumscribed rectangle having a minute or excessive size from the circumscribed rectangles created by the circumscribed rectangle processing section 506. This size is given as a pixel value such as, for example, 8 × 8 (pixels) or less as the size of a minute circumscribed rectangle and 120 × 120 (pixels) or more as the size of an excessively large circumscribed rectangle. FIG. 9 shows an example of a processing area designating method in the processing area designating section 107. The designation method, the entire region excluding the entire image on the postal matter P, and specific region 1201 1
204 (FIG. 9a), relative area 1 from specific area 1201
202, 1203 (FIGS. 9b and 9c), and one of these is designated. FIG. 10 shows the configuration of the clustering processing section 109 in detail. Clustering processing unit 109
Then, the data blocked by the blocking processing unit 108 is integrated and clustered (hereinafter, the clustered cluster portion is also referred to as a row). In the present embodiment, the direction in which blocks are integrated is limited to two directions, the horizontal direction and the vertical direction, but the integration direction is not limited to the horizontal direction and the vertical direction. In the clustering processing described below, processing is performed only on the processing area specified by the processing area specifying unit 107. The horizontal integration processing unit 1301 is configured as shown in FIG.
The image data divided into blocks as shown in FIG. 11A is integrated (clustered) in the horizontal direction as shown in FIG. An example of the integration method in the horizontal direction will be described with reference to the integration of the blocks 1401 and 1402 in FIG.
FIG. 11C shows a block 140 in FIG.
1, 1402 is extracted, but block 1401 is shown.
From the viewpoint of integrating the block 1402,
Specify as 04. If at least one block 1402 is included in this area, blocks 1401 and 1402
And integrate. However, here, as shown in FIG. 11D, when the block 1405 protrudes from the upper end of the integration range 1404 of the block 1401, it is determined that no integration is performed. The threshold of the integrated range in FIG. 11C is given as, for example, a = 6 (pixels), b = 35 (pixels), c = 6 (pixels), and d = 35 (pixels). By the above processing, the block data shown in FIG. 11A is integrated into three clusters (rows) 1403 as shown in FIG. 11B. In addition, the vertical direction integration processing unit 1302 integrates the blocked data as shown in FIG. 12A in the vertical direction by the same method as described above. By such a vertical integration process, in this example, integration into clusters of five rows is performed as shown in FIG. FIG. 13 shows the configuration of the cluster number counting / comparing section 110 in detail. In the cluster number counting / comparing section 110, first, the horizontal row counter 1601 and the vertical counter 1602 count the number of rows (the number of clusters) in each integration direction. In the examples of FIGS. 11 and 12, the number of integrated rows in the horizontal direction is “3”, and the number of integrated rows in the vertical direction is “5”. Next, the line number comparison unit 1603 outputs the direction having the smallest number of lines among the integration directions as the writing direction of the character line. In this example, the horizontal direction is output as the writing direction of the character line. This output result is the writing direction of the address character line on the mail P. As described above, according to the above-described embodiment, since the determination of the writing direction of the address character line on the postal matter is determined on the basis of the characteristic described in the address character line, the conventional post-pay charge method is used. It is possible to determine the writing direction of the address character line regardless of the direction of the line inside the round frame of the mark or the separate mark or the typeface of the character, and even if they do not exist on the mail. Since the writing direction is determined based on the characteristics of the entire address character line instead of analyzing each character, the writing direction of the address character line can be determined regardless of the determination result of each character. It is. Further, since the writing direction of the address character line of the postal matter can be determined, the recognition direction can be specified at the time of character recognition, and it is not necessary to perform character recognition assuming another writing direction. In the above-described embodiment, a case has been described in which the address reading device for a postal matter determines the writing direction of the address character line that constitutes the address information described on the postal matter. The present invention is not limited to this. For example, the present invention can be similarly applied to a case where the direction of a character line is determined in a character reading device or the like that reads characters described in other forms or securities. As described above in detail, according to the present invention, a character line direction discrimination that can accurately discriminate the direction of a character line on a read object without being restricted by the conventional one. Equipment can be provided.

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a block diagram showing a configuration of a mail address reading device to which a character line direction discriminating device according to the present invention is applied. FIG. 2 is a block diagram showing a configuration of a blocking processing unit in detail. FIG. 3 is a diagram for explaining a differentiation process of a differentiation processing unit. FIG. 4 is a view for explaining the same encoding processing of the same encoding processing unit. FIG. 5 is a diagram illustrating a local maximum point detection process performed by a local maximum point detection processing unit; FIG. 6 is a view for explaining an isolated point removal process of an isolated point removal processing unit. FIG. 7 is a diagram illustrating a labeling process performed by a labeling processing unit. FIG. 8 is a diagram for explaining a circumscribed rectangle process performed by a circumscribed rectangle process unit; FIG. 9 is a view for explaining an example of a processing area designation method in a processing area designation unit. FIG. 10 is a block diagram showing the configuration of a clustering processing unit in detail. FIG. 11 is a view for explaining horizontal integration processing of a horizontal integration processing unit. FIG. 12 is a view for explaining vertical integration processing of a vertical integration processing unit. FIG. 13 is a block diagram showing the configuration of a cluster number counting / comparing unit in detail. [Description of Signs] P: Mail (reading material), 101: Light source, 102:
Optical system, 103: line sensor, 104: image input unit (image input means), 105: A / D conversion unit, 106
...... image memory 107 processing area designation unit 108
... Blocking processing unit, 109 ... Clustering processing unit, 1
10 cluster count / comparison unit 111 character recognition unit 501 differentiation processing unit 502 encoding processing unit 503 local maximum point detection processing unit 504 isolated point removal processing unit , 505 ... Labeling processing unit, 506 ...
Circumscribing rectangle processing section, 507 minute / excess label removal processing section, 1301 horizontal integration processing section, 1302 vertical integration processing section, 1601 horizontal row counter section, 1602 vertical Direction counter section, 1603... Row number comparing section.

Claims (1)

(57) [Claims] 1. An image input means for inputting an image including a character line on a reading material in which a character line described in a horizontal direction or a vertical direction exists, and the image input means differential Mato on the input image by
Differentiate horizontally and vertically using Rix
Differentiating means and each of the differentiated in the horizontal and vertical directions by the differentiating means
By performing the process of taking the absolute value of the pixel value,
Same encoding means for matching the sign of the value, and
A local maximum point that detects the local maximum point from each pixel value in the vertical direction
Detecting means, and the horizontal and vertical directions detected by the maximum point detecting means.
Binarization by taking the logical sum of the local maximum points
Means for each pixel with respect to the image binarized by the binarizing means.
When the pixel shows a value isolated from the surrounding pixels
The isolated point is removed by replacing that pixel with a specific value.
Means for removing isolated points, and the effectiveness of the image from which isolated points have been removed by the isolated point removing means.
Focusing on the pixel, if there is an effective pixel in the vicinity, the image
By performing the processing to make the elements the same label,
And labeling means for bridging, to the labeled image by the labeling unit
By creating the smallest circumscribed rectangle containing the image
Circumscribing rectangle means for making a circumscribing rectangle, and an image circumscribed by the circumscribing rectangle means.
A small or excessively small circumscribed rectangle is removed.
Excessive circumscribed rectangle removing means and the result of processing by the minute / excessive circumscribed rectangle removing means
The circumscribed rectangularized image horizontally and vertically
And integrating means for integrating by clustering, respectively, counting means for counting each integrated each direction integration number This integration means, said reading Butsujo based on each integrated number counted by the counting means A character line direction determination device, comprising: a determination unit configured to determine a direction of a character line.