JPH0324681A - Character recognizing device - Google Patents

Abstract

PURPOSE:To segment an inclined character string by distinguishing the image data of a character line in the main scanning direction, extracting a character existence area, detecting character projection information in the subscanning direction, segmenting a character area, and executing character recognition processing. CONSTITUTION:The image data of a character string from an original image obtained from a reader 20 are inputted through an input part 2 and stored in a RAM 7. A CPU 5 divides the image data into 4 equal parts in the line direction and checks a scanning line number including the image data in each division. The image data of character strings other than the target string are removed and the projection area of each division of the target character string is extracted. The projected image of the character line in the height direction is found out. Character width in the projection area is determined and the character area of one character is segmented from the height and width to recognize the character. Said processing is applied to all characters.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a character recognition device, and for example, to a character recognition device that cuts out characters from a diagonal character string.

[Prior Art] Conventionally, in this type of device, for characters written horizontally (or vertically), the projection in the horizontal (or vertical) direction is taken out to cut out the lines, and then the vertical (or vertical) Alternatively, characters are extracted by extracting the projection in the horizontal direction.

An example of this will be explained using FIGS. 5(a) and 5(b).

First, in the same figure, ℃ is the line length, h゜ is the line spacing, X is the slope limit angle of line direction projection, h is the character height, β゜ is the character spacing, and y is the slope in the character height direction. Indicates the limit angle, and using the above symbols, the following formula holds true for Xl3'. That is, h x=tan-' β (1) β y=tan-' - (2)
h.

[Problems to be Solved by the Invention] However, in the above conventional example, when cutting out oblique characters, as shown in FIGS. 5(a) and (b),
Since the ratio β:h゜ between the line length (I2) and the line spacing (h゜) is very large compared to the ratio h:β゛ between the character height h and the character spacing 2゛, the line projection is Since the limit angle of slope, The present invention has been made in view of the above-mentioned drawbacks of the conventional example, and its object is to provide a character recognition device that can cut out characters even in a character string with a large inclination in a document.

[Means for Solving the Problems] In order to solve the above-mentioned problems and achieve the objectives, a character recognition device according to the present invention includes an input means for inputting image data containing at least l lines of character line information; a dividing means for dividing the image data inputted by the input means in the main scanning direction; an extracting means for extracting an area where the character line information exists for each image data divided by the dividing means; a detection means for detecting character projection information in the sub-scanning direction in each extracted region; a character cutting means for cutting out a character region based on the character projection information detected by the detection means; and recognition means for performing character recognition processing for each character area cut out by the character cutting means.

[Operation] According to this configuration, the input means inputs image data including at least l lines of character line information, the sorting means sorts the image data inputted by the input means in the main scanning direction, and the extracting means The segmentation means extracts a region where character line information exists for each segmented image data, and the detection means detects character projection information in the sub-scanning direction in the region for each segment extracted by the extraction means, and performs character cutting. The output means cuts out a character area based on the projection information of the character detected by the detection means, and the recognition means performs character recognition processing for each character area cut out by the character cutting means.

[Embodiments] Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings.

FIG. 1 is a block diagram illustrating the configuration of the character recognition device of this embodiment. In the figure, 1 indicates the character recognition device of this embodiment, 20 indicates a reading device that sends image data optically read from a document to this device 1, and 30 indicates character recognition recognized by this device 1. A display device for inputting and displaying results is shown. Note that the output destination of the recognition results is the display device 3.
0 may be replaced by a personal computer.

In the apparatus 1, numeral 2 indicates an input section for inputting dot data from the reading device 20, and numeral 8 indicates an output section for outputting recognition results, ie, data, to the display device 30. Reference numeral 3 indicates an operation section for performing operations such as reading a manuscript and displaying character recognition results. Reference numeral 4 indicates a character recognition unit that performs character recognition in an area where characters have been cut out from read image data, and 5 indicates an RO unit that controls the overall operation of the device 1.
A CPU is shown which controls using various programs in the MG, and 6 shows a ROM which stores a control program, an error processing program, a program of the flowchart shown in FIG. 2, and the like. Reference numeral 7 indicates a RAM used as a work area for the various programs mentioned above and a temporary save area during error processing. Reference numeral 9 indicates a bus line for transmitting data, address signals, and control signals within the device 1.

Next, the operation of this embodiment will be explained.

FIG. 2 is a flowchart illustrating the character recognition operation of this embodiment, and FIG. 3 is a diagram illustrating the process up to cutting out an area for one character.

First, image data of a document image read by the reading device 20 is input via the input unit 2 and then stored in the RAM 7 (step SL). Then, the first step (1
), A-D in the row direction for the image data.
The image is divided into four equal parts (step S2). According to FIG. 3, the middle character line is the character line to be cut out.

To check the existence of image data for the target character line,
Scanning lines that divide the sub-operation direction (No. 1 to
line number up to 12 is assigned), the scanning line number where image data exists for each section is determined. is examined as in the second step (II) of FIG. In this way, as shown in the second step (n) of FIG.
The image data of each target character string divided into equal parts is classified into division A.
6th line to 12th line, and 4th line for section B.
It is confirmed that there are lines from line 1 to line 10, line C from line 3 to line 10, and line D from line 2 to line 8. .

After this, the image data of character lines other than the target character line are removed, and as shown in the third step (III) in FIG.
,B,C,D. First, region A is extracted (step s3), and as shown in the third step (III) in FIG. 3, the projection in the height direction of the character line is obtained (
Step S5). The processing of steps S4 and s5 is repeated until areas A-D (step S6
).

Next, characters are cut out. For example, as shown in the fourth step (■) in FIG. S belonging to
Since the projection indicated by l and the projection indicated by S2 belonging to area B are continuous, the character areas indicated by CI and C2 of the image data corresponding to each projection are combined, and thereby the height h of the character and the character The width W of the character is determined, and a rectangular area (character) formed by the height h of the character and the width W of the character is cut out as one character (step S7). and,
Character recognition of the cut out characters is performed (step S8),
The recognition result is sent to the display unit 3o (step S9)
. In this way, characters are repeatedly cut out from the currently targeted character line between step 87 and step s9 (step SIO). Furthermore, steps 83 to 83 until character recognition of all character strings in the original image are completed
The process of step SIO is repeated (step SIO).
ll) As explained above, according to this embodiment, characters can be reliably extracted even from a skewed character string.

Now, in the above-mentioned embodiment, characters are cut out by determining the projection in the row direction (main scanning direction) and the height direction (sub-scanning direction), but the present invention is not limited to this. Instead, as a modification of the above-described embodiment, a method may be used in which, especially when the inclination of the character string is large, the optimum inclination of the character string is determined and the characters are extracted. Below, a modification thereof will be explained.

FIG. 4 is a diagram illustrating a modification.

In this modified example, as shown in FIG. 4, the image data of the read original image is divided into four equal parts A-D in the same way as in the above embodiment, and by projection in the row direction, the character string to be recognized is Once this is determined, the slope X of the character string is then calculated. The method for calculating the slope X is a well-known technique, so the explanation thereof will be omitted.

Then, based on the slope X, as shown in FIG.
Projection in the height direction of the character is taken, and as in the above-described embodiment, in successive projections, one rectangular area is determined from the character area corresponding to each projection, and that rectangular area is cut out and used for character recognition. will be implemented.

According to this modification, even if the character string has such an inclination that the projections in the height direction overlap between adjacent characters, the characters can be accurately extracted.

In this modified example, if the character strings in each area have a constant slope, there is no need to check the slope for each character string, and the slope calculated for the first line of characters is used for all characters. All you have to do is cut out the characters from the column.

[Effects of the Invention] As described above, according to the present invention, characters can be reliably extracted even from a skewed character string.

[Brief explanation of drawings]

Figure 1 is a block diagram explaining the configuration of the character recognition device of this embodiment, Figure 2 is a flowchart explaining the character recognition operation of this embodiment, and Figure 3 is the process of cutting out an area for one character. 4 is a diagram illustrating a modified example, and FIGS. 5(a) and 5(b) are diagrams illustrating a conventional character cutting method. In the figure, l...Character recognition device, 2...Input section, 3...
・Operation unit, 4...Character recognition unit, 5...CPU,...
- ROM, 20...Reading device,...RAM, 8...Output 30...Display device.

Claims (2)

[Claims] (1) An input means for inputting image data including at least one line of character line information, a sorting means for sorting the image data inputted by the input means in the main scanning direction, and an image divided by the sorting means. an extraction means for extracting a region where the character line information exists for each data; a detection means for detecting character projection information in the sub-scanning direction for each region extracted by the extraction means; The present invention is characterized by comprising a character cutting means for cutting out a character area based on projection information of a character, and a recognition means for performing character recognition processing for each character area cut out by the character cutting means. Character recognition device. (2) an input means for inputting image data including at least one line of character line information; a sorting means for sorting the image data inputted by the input means substantially in the main scanning direction; Extracting means for extracting an area where the character line information exists for each image data; Tilt detection means for detecting the tilt of the character line with respect to the main scanning direction based on the area extracted by the extracting means; and the tilt detection. projection detection means for detecting projection information of characters in the regions of the respective divisions based on the inclinations detected by the means; and cutting out character regions based on the projection information of characters detected by the projection detection means. A character recognition device comprising: a character cutting means; and a recognition means for performing character recognition processing for each character area cut out by the character cutting means.