JPH05182024A - Character recognition device - Google Patents

Abstract

PURPOSE:To segment a line in a document picture where space between the lines is narrow and a document picture which is inputted with inclination. CONSTITUTION:A controller 1 divides inputted document picture data so as to be perpendicular to a line direction in accordance with the segment control program of segment control program ROM 5, and takes projecting data for respective pieces of divided document picture data. The controller 1 extracts document areas for respective divided areas from projecting data and fetches a line area in inputted document picture data from the document areas.

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,
The present invention relates to a character recognition device that can accurately perform line segmentation and perform character recognition even when the line spacing of a document image is narrow or the document image is inclined.

[0002]

2. Description of the Related Art In a conventional character recognition device, when character recognition is performed, as a line segmentation method, density projection in a line direction (horizontal direction for horizontal writing, vertical direction for vertical writing) is performed on a document image. A method is known in which data is taken and the range of data is 0 (or the range of data is less than or equal to a threshold value) is used as the space between lines.

[0003]

However, according to the above-mentioned method, when the line spacing of the document image is narrow (when the lines are in contact with each other or when the line area is digging) or when the document image is tilted, one line of the document image is displayed. There is a problem that it cannot be cut out accurately.

In view of the above conventional problems, an object of the present invention is to provide a high-performance character recognition device capable of cutting out a line even when the line spacing of the document image is narrow or the document image is inclined. It is in.

[0005]

A character recognition device of the present invention comprises a control device 1 and a cutout control program ROM 5 as a dividing device, a control device 1 and a cutout control program ROM 5 as a data extracting device, and a document area extracting device. The control device 1 and the cutout control program ROM 5 as the above, and the control device 1 and the cutout control program ROM 5 as the detection means are provided.

[0006]

In the character recognizing device having the above-mentioned structure, the control device 1 cuts out the line of the document image data inputted by the input means in the following manner in accordance with the cut-out control program stored in the cut-out control program ROM 5. That is, the control device 1
The document image data input by the input means is divided into the row direction and the vertical direction, and projection data is obtained for each divided document image data. Then, the control device 1 extracts the document area for each divided area from the projection data. Further, the control device 1 extracts the line area in the input document image data from the document area for each divided area.

Next, control device 1 performs character recognition processing on the fetched line area. That is, the control device 1 cuts out a character for each line area and recognizes a character image for each line.

By doing so, even if the line spacing of the document image is narrow or the document image is tilted, line cutting can be accurately performed, and a high-performance character recognition device can be provided.

[0009]

Embodiments of the present invention will now be described with reference to the drawings. FIG. 1 is a block diagram showing an embodiment of a character recognition device according to the present invention. In the figure, reference numeral 1 is a control device, and the control device 1 communicates with a communication control device 3 via a bus 2.
RAM 4 for calculating control device, cutting control program RO
M5, recognition program ROM6, image storage RAM7,
It is connected to the dictionary storage ROM 8. Here, the communication control device 3 receives the document image from the input means (not shown),
The recognition result is output. The cutout control program ROM 5 is an R storing the cutout control program.
OM. The recognition program ROM 6 is a ROM that stores a recognition program. The image storage RAM 7 is
RA for saving input document images, single-line character images, etc.
It is M. The dictionary storage ROM 8 is a ROM that stores a dictionary for character recognition.

Next, the overall operation will be described with reference to FIG. Note that FIG. 2 is an overall operation flowchart showing an embodiment of the character recognition device of the present invention. Control device 1
Captures the document image data input from the communication control device 3 and stores it in the image storage RAM 7. The control device 1 cuts out lines from the input document image data stored in the image storage RAM 7 according to the cutout control program in the cutout control program ROM 5 (step S).
1).

The control device 1 performs the character recognition process on a certain cut-out line area as follows. That is, the control device 1 checks whether or not the height of a line in a certain line region that has been cut out is too large. If it is too large, the next line is cut out (steps S2, S6, S1), and if it is not too large Performs in-line recognition (step S3). Step S3
Then, the control device 1 cuts out a character in a certain line area cut out in accordance with the cutout control program stored in the cutout control program ROM 5, and then the recognition program ROM 6
The character image is recognized in accordance with the recognition program of 1. and using the dictionary of the dictionary storage ROM 8. Next, the control device 1 performs predetermined post-processing (step S4), and transmits the recognition result to the display device via the communication control device 3 and displays it (step S5). Then, the next line is cut out (steps S6 and S1).

Thereafter, the same operation as described above is repeated for all the lines of the input document image data (steps S1 to S1).
S6).

Next, the line cutout process of step S1 of FIG. 2 will be described with reference to FIG. FIG. 3 is an operation flowchart of a main part showing an embodiment of the character recognition device of the present invention, and particularly an operation flowchart showing an embodiment of line cutting out in step S1 of FIG.

The document image data is supplied to the communication control device 3 by the input means. The control device 1 takes in the document image data from the communication control device 3 and stores it in the image storage RAM 4. The control device 1 cuts lines from the document image data as follows according to the cut-out control program stored in the cut-out control program ROM 5.

That is, the control device 1 divides the document image data stored in the image storage RAM 4 perpendicularly to the row direction, and takes projection data in the row direction for each divided area (step S1).
1). The control device 1 extracts a document area (hereinafter referred to as a unit) for each divided area from the obtained projection data (step S12). The control device 1 has the following divided areas,
Label overlapping units with the same label. At that time, if the overlap is very small, the same label is not used (step S13). Next, the control device 1 determines whether or not line contact has occurred for each label from the characteristics of a set of units (hereinafter referred to as a unit group). If line contact has occurred, the height of the document area of each divided area is increased. From this, the place where the line contact is occurring is specified (step S14).

As a result of the determination, if there is no line contact (if all lines are one line), the control device 1 starts isolated point processing (steps S15 and S23). If there is contact (if not all in one line),
A disconnection process is performed on the units that are in row contact (steps S15 and S16). Here, the number of lines of the document area in which the unit group causing line contact is estimated is estimated, and the unit to be cut, the number of cuts, and the cutting position are determined.

Next, if there is a disconnected unit, the control device 1 performs the labeling process again (steps S17 and 18), and if there is no disconnected unit, the control device 1 enters the forced line cutting process (step S17,). S2
1). When the control device 1 has entered the labeling process of step S18, it determines whether line contact has occurred (step S19), and when line contact has occurred, it enters forced line cutout processing (steps S20 and S21). If there is no contact, isolated point processing starts (steps S20 and S2).
3).

In the forced line cut-out process of step S21, the control device 1 estimates the number of lines of line contact, and separates the unit group according to the estimated number of lines. After this,
After determining whether or not row contact has occurred, the control device 1 starts isolated point processing (steps S22 and S23). In the process before reaching the isolated point process (step S23), the upper and lower separated characters may be determined as different lines. Therefore, in the isolated point process (step S23), the control device 1 detects an isolated point such as noise. It is determined whether or not it is included in the surrounding lines.

Further, the control device 1 judges from the characteristics of the unit group whether line contact has occurred for each label (step S24) and creates a line cutout list (step S25).

Next, a specific example of the result of processing the document image (FIGS. 4 to 8) will be described. 4 to 8 are examples of the result of processing the document image. In these figures, the portion of the rectangle 12 that is shaded is a document area for each divided area 11.

FIG. 4 shows an example of a document image which is input horizontally with a sufficient space between lines. FIG. 5 is an example of a document image in which lines are input with a tilt, but the line spacing is sufficient. In these cases of FIG. 4 and FIG. 5, the line segmentation is completed (steps S23 to S25) in the first labeling process (steps S13, S14, S15) in FIG.

FIG. 6 shows an example of a document image in which lines are input horizontally but the line spacing is narrow and line contact occurs.
In the case of FIG. 6, after the first labeling process (step S13) in FIG. 3, the presence or absence of row contact is determined (steps S14 and S15), and the second divided area 11 from the left is displayed.
The document area 12 is cut (step S16), and the second labeling process is performed (steps S17 to S2).
0), the line segmentation processing is terminated (steps S23 to S).
25).

Further, FIG. 7 shows an example of a document image in which line contact is caused because the line spacing is narrow and the input is performed with an inclination.
In the case of FIG. 7, row contact occurs after the first labeling process (step S13) in FIG. 3 (FIG. 7).
However, since there is no document area 12 to be cut, forced line cutout processing is started (steps S15 to S1).
7, S21).

FIG. 8 shows an example of a document image in which the upper points such as i and j do not contact the original line area. In the case of FIG. 8, steps S11 to S15 in FIG.
The isolated point process of step S23 is performed, and the line segmenting process is completed (steps S24 and S25). Here, in the isolated point processing of step 23, the leftmost divided area 1 in FIG.
If there is a sufficiently large document area 12 below the isolated point 13 like the isolated point 13 at 1, the line is determined to be the same as the document area 12. If there is no document area below the isolated point 13, such as the isolated point 13 in the second divided area 11 from the right in FIG.

The present invention is not limited to this embodiment, and various applications and modifications are conceivable without departing from the gist of the present invention.

[0026]

As described above, according to the character recognition apparatus of the present invention, when cutting out a line in the case of performing character recognition of a document image, the line cutting means narrows the line spacing (contact or bite). It is possible to perform line segmentation accurately even for a document image or a document image that is input with a tilt, reduce constraints on the input document, and adapt to various documents.

[Brief description of drawings]

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

FIG. 2 is an overall operation flowchart showing an embodiment of a character recognition device of the present invention.

FIG. 3 is a main part operation flowchart showing an embodiment of a character recognition device of the present invention.

FIG. 4 is a diagram showing an example of a result of processing a document image.

FIG. 5 is a diagram showing another example of a result of processing a document image.

FIG. 6 is a diagram showing another example of a result of processing a document image.

FIG. 7 is a diagram showing another example of a result of processing a document image.

FIG. 8 is a diagram showing another example of a result of processing a document image.

[Explanation of symbols]

 1 Control device 5 Cutting control program ROM 6 Recognition program ROM 8 Dictionary storage ROM

Claims (1)

[Claims] 1. A character recognizing device comprising: a line slicing device for extracting a line region from document image data input by an input device; and a character recognizing device for recognizing a line region extracted by the line slicing device. The line cutting-out unit divides the document image data input by the input unit into a line direction and a vertical direction, a data extraction unit that takes projection data for each document image data divided by the dividing unit, and the data. Document area extracting means for extracting a document area for each divided area from the projection data obtained by the extracting means, and detecting means for extracting a line area in the input document image data from the document area for each divided area extracted by the document area extracting means A character recognition device comprising: