JPH0573718A - Area attribute identifying system - Google Patents

Abstract

PURPOSE:To decide a character assembling direction even for a document with narrow character gap and to automatically perform character recognition by detecting the connecting direction of a character from distance information between circumscribing rectangles of connection of black picture elements. CONSTITUTION:A character assembling direction decision part 10 checks the connecting states of all the black picture elements in an area decided as a character area, and detects the coordinates of the rectangular circumscribing to a cluster in which the black picture elements are connected. A circumscribing rectangle nearest to the circumscribing rectangle is detected for each circumscribing rectangle, and the position relation of it is detected. When it is connected in the horizontal direction, a horizontal direction coupling counter is counted by one, and when it is connected in the vertical direction, a vertical direction coupling counter is counted by one. When the processing of all the circumscribing rectangulars are completed, the value of the horizontal direction coupling counter is compared with that of the vertical direction coupling counter, and it is decided that the area is a horizontal writing area when the value of the horizontal direction coupling counter is higher, and that it is a vertical writing area when the value of the vertical direction coupling counter is higher.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a preprocessing of a character recognition device capable of recognizing a print document in which characters and / or graphics are mixed according to the attributes of areas such as characters and / or graphics.

[0002]

2. Description of the Related Art A conventional area attribute identification method will be described below. First, the area is divided based on the white pixel information of the binary data captured by the scanner. Then, it is determined whether the area divided from the horizontal / vertical histogram of black pixels and the circumscribed rectangle is a character area, a graphic area, or a photographic area. If the determined area is a figure, vectorization of the figure is performed, and if it is a photographic area, data compression processing is performed on the image area. If the determined area is a character area, it is further determined whether the writing direction of the character is a vertical writing area or a horizontal writing area, and character recognition is performed through character cutting processing according to the vertical writing document / horizontal writing document. Perform processing.

The following processes are performed in the conventional determination of the character set direction. Histograms in the horizontal and vertical directions are taken for the area determined to be the character area, the peak value of each histogram is detected, and the peak distance is obtained. This distance is character size + line spacing or character size +
Since it is between characters and the line spacing is larger than the character spacing, the average of the peak-to-peak distances of the histogram in the horizontal direction and the average value in the vertical direction are compared, and the smaller direction is defined as the character-set direction.

[0004]

In the conventional area attribute identification method, when a histogram is taken for a document such as a newspaper where the character spacing is narrow, the histogram in the direction other than the character set direction has a gentle shape. Therefore, it is difficult to accurately detect the peak value of the histogram, and there is a problem that it is not possible to correctly detect the direction of a character set.

The present invention has been made to solve the above-mentioned problems, and has been made with the object of providing an area attribute identification method capable of determining the direction of a character set even for a document having a narrow character space. is there.

[0006]

SUMMARY OF THE INVENTION In order to solve the above problems, the present invention performs character segmentation in which a binarized recognition target document is divided into regions, the attributes of the divided regions are identified, and recognition is performed. In the apparatus, means for converting a recognized document into binary data, means for dividing an area from white pixel information of the binary data, and identification of an attribute of the area from black pixel information in the area for each divided area. And means for detecting the connecting direction of the characters from the distance information between the circumscribing rectangles of the black pixel connection when the identification result is the character area, and the character recognition is performed by automatically detecting the character setting direction.

[0007]

According to the present invention, with the above configuration, the attributes of the divided areas are identified, and the area determined to be the character area is
Check the connection state of all black pixels in the area, detect the coordinates of the rectangle circumscribing the lump where the black pixels are connected,
Store the circumscribed rectangles in ascending order.

Next, the following processing is performed in order on the circumscribed rectangles stored in ascending order. (Because this circumscribed rectangle is a base point for detecting the distance from other rectangles, it will be referred to as a base point rectangle hereinafter.) The range in which the rectangle having the closest distance from the base point rectangle is searched is determined for each base point rectangle. A rectangle within that range (below,
The distance between the target rectangle) and the base rectangle is calculated, and the rectangle with the shortest distance is determined. The positional relationship between the determined minimum distance rectangle and the base point rectangle is examined, and if it is located in the horizontal direction, one horizontal coupling counter is recommended. Also, if it is located in the vertical direction, one vertical coupling counter is recommended.

When the processing is completed for all the circumscribed rectangles, the values of the horizontal direction coupling counter and the vertical direction coupling counter are compared. If there are many horizontal coupling counters, it is determined to be a horizontal writing area, and if there are many vertical coupling counters, it is determined to be a vertical writing area.

[0010]

An embodiment of the present invention will be described with reference to the drawings. FIG. 1 shows a device block for performing area attribute identification in one embodiment of the present invention. In FIG. 1, reference numeral 1 is a ROM in which an area attribute identification program is stored. Reference numeral 2 denotes a CPU for identifying a region attribute, which includes a region dividing unit 6, a region determining unit 7, a graphic vectorization processing unit 8, an image compressing unit 9, a character framing direction determining unit 10 in FIG.
It has a character cutout unit 11 and a character recognition unit 12. 3 is an R for storing binary data captured by the scanner 4
AM. A display device 5 displays the recognition result of the area identified by the CPU.

In the area dividing unit 6 there are 2 documents to be recognized.
The value data is fetched from the scanner 4 and stored in the RAM 3. The horizontal blank band and the vertical blank band are detected from the captured binary data, and the area surrounded by the detected blank band is defined as one area, and the coordinate values of the upper left and lower right of the area are set. Store in RAM3.

The area determination unit 7 determines whether the divided area is a photographic area, a graphic area, or another area. For the area determined to be the character area, the character embedding direction determination unit 1
The character set direction is determined by 0, the character cutout unit 11 performs character cutout processing according to the character set direction, and the character recognition unit 12 performs recognition processing.

The determination of the direction of the letter combination will be described below with reference to the flow chart of FIG. 3 and the region examples of FIGS. 4 and 5. The connection state of all the black pixels in the character area is checked, and the value of (x1, y1, x2, y2) of the coordinates of the rectangle circumscribing the block in which the black pixels are connected is detected (see FIG. 4).
), And the y1 of the circumscribing rectangle is stored in ascending order (steps 1 and 2). In the example of FIG. 5B, the circumscribed rectangle 1 has coordinates (4, 4, 14, 12) and the circumscribed rectangle 2 has coordinates (21,
4, 31, 16), for example, rectangle 1, rectangle 2, ...
.. are stored in the order of rectangle 6 (however, since y1 is stored in ascending order, rectangle 1 and rectangle 2 may be reversed). Also, a horizontal combination counter y for determining the character direction.
ok, the vertical combination counter tate is cleared to 0 (step 3).

First, the first circumscribed rectangle stored in ascending order is set as a base point rectangle (step 4). The range of the target rectangle in which the closest distance from the base point rectangle is searched is when the value of y1 of the target rectangle is smaller than the following value.

Y2 of the base point rectangle + 1/2 of the shorter width and height of the base point rectangle The width of the rectangle 1 in FIG. 5B is 11 (14-4 + 1) and the height is 9 (12-4 + 1). So the search range is 12 + 9/2
= 16, and the rectangles whose y1 values are smaller than 16 are rectangle 2 and rectangle 3. Therefore, there are two search target rectangles, rectangle 2 and rectangle 3, when rectangle 1 is the base rectangle (step 5).

In order to detect the minimum distance rectangle, the values of the minimum distance rectangle minBox and the minimum distance minDis are cleared (step 6). It is checked whether the value of y1 of the target rectangle is within the range, and if it is within the range, the following processing is performed. If not, it is considered that the distance search for one base point rectangle is completed, and the process proceeds to step 14. To jump.
When the rectangle 1 is the base rectangle, the process jumps when the target rectangle becomes the rectangle 4 (step 7).

Next, the value of y1 of the target rectangle and y2 of the base point rectangle
Are compared (step 8). If the value of the target rectangle y1 is larger than the base point rectangle y2, the process goes to the check process for vertical direction connection (step 9), and if not, the process goes to the check process for horizontal direction connection (step 11). Rectangle 2 advances to the horizontal processing unit, and rectangle 3 advances to the vertical processing unit.

At (step 9), if the following conditions are satisfied, it becomes a candidate for the minimum distance rectangle. If the ratio of the width of the target rectangle to the width of the base rectangle is 2 or less and the rectangles overlap each other in the vertical direction, they become candidates for the minimum distance rectangle, and the process proceeds to step 10. If the value of y1 of the target rectangle is the inter-rectangular distance and the distance obtained here is smaller than the minimum distance minDis, the minimum distance minDis is set to this distance (inter-rectangle distance). Also, the minimum distance rectangle minB
Information of this target rectangle is set in ox.

In (step 11), a candidate for a minimum distance rectangle is obtained if the following conditions are satisfied. The ratio of the height of the target rectangle and the height of the base rectangle is 2 or less, and the value of x2 of the target rectangle is x1-1 / 2 of the width and height of the base rectangle of the base rectangle x1.
It is larger and the value of x1 of the target rectangle is x of the base rectangle
If it is smaller than ½ of the shorter width and height of the 2 + base point rectangle, it becomes a candidate for the minimum distance rectangle and proceeds to step 12. Note that step 11 considers that the rectangle 1 and the rectangle 2 may be stored in reverse. When the target rectangle is located to the left of the base rectangle, the value of x1 of the base rectangle is the inter-rectangle distance, and when the target rectangle is located to the right of the base rectangle, x1 of the target rectangle. Base point x
The value of 2 is the distance between the rectangles. If the distance obtained here is smaller than the minimum distance minDis, the minimum distance minDis
The distance here is set. The information of the target rectangle is set in the minimum distance rectangle mimBox.

If the minimum distance rectangle detected by the above processing is located in the horizontal direction with respect to the base point rectangle, then one horizontal direction coupling counter yoko is recommended, and if it is located in the vertical direction, it is vertical. One directional coupling counter tate is recommended (steps 14-16). The above processing is continued until there is no base point rectangle (steps 17 and 1).
8).

Finally, when all the circumscribed rectangles have been processed, the values of the horizontal direction coupling counter yoko and the vertical direction coupling counter tate are compared. When the horizontal combination counter yoko is large, it is determined to be a horizontal writing area, and when the vertical combination counter tate is large, it is determined to be a vertical writing area (steps 19 to 21).

In the example of one area of FIG. 5, the minimum distance rectangle of rectangle 1 ... rectangle 3 the minimum distance rectangle of rectangle 2 ... rectangle 5 the minimum distance rectangle of rectangle 3 ... none the minimum distance rectangle of rectangle 4 .. None. Minimum distance rectangle of rectangle 5 ... None. Minimum distance rectangle of rectangle 6 ... None. Horizontal counter yoko = 0, vertical counter ta.
The result of te = 2 is obtained, and this area is determined to be a vertically set area.

[0023]

According to the present method, the character set direction is determined by paying attention to the connecting direction of the characters in the character area. Therefore, it is impossible to detect the character set direction by the histogram. With respect to, it is also possible to determine the direction of character combination, and it is possible to automatically perform character recognition.

Further, since the rectangles are sorted in ascending order so as not to be compared with all the rectangles in the area and the search range is determined for each rectangle, the processing can be performed at high speed. Has the effect.

[Brief description of drawings]

FIG. 1 is a block diagram of an apparatus for performing area identification according to an embodiment of the present invention.

FIG. 2 is a block diagram of an embodiment of the present invention.

FIG. 3 is a flowchart in the embodiment of the present invention.

FIG. 4 is a flowchart in the embodiment of the present invention.

FIG. 5 is a diagram showing coordinates of a circumscribed rectangle.

FIG. 6 is a diagram showing a character area.

[Explanation of symbols]

 1 ROM 2 CPU 3 RAM 4 Scanner 5 Display 6 Area Division 7 Area Judgment Section 10 Character Set Direction Determining Section

Claims (1)

[Claims] 1. A character recognition device for dividing an area of a binarized recognition target document, identifying attributes of the divided area, and recognizing the divided document, and means for converting the recognized document into binary data. A means for dividing the area from the white pixel information of the binary data, a means for identifying the attribute of the area from the black pixel information in the area for each divided area, and a connecting direction of characters when the identification result is a character area. An area attribute identification method that includes means for detecting the distance from the circumscribing rectangles connected to black pixels, and automatically detects the character set direction to perform character recognition.