JP3039427B2 - Character extraction method and method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a character extracting technique in an optical character reading apparatus, and more particularly, to a character extracting method and a character extracting method for characters in contact with each other.

[0002]

2. Description of the Related Art The character segmentation technique is used by an optical character reader or the like to extract a character to be read from an image captured and digitized by an image input method such as optical scanning. As a conventional character segmentation method, for example, Japanese Patent Application Laid-Open No. 6-203201 discloses that character segmentation candidates are digitized in order to enable high-performance recognition of a character string in which adjacent characters approach and overlap with a small amount of processing. Is evaluated by performing labeling processing on the image, and assuming that the cut-out candidates are biased or made up of kanji, an evaluation function indicating the character-likeness (validity) when combining the character cut-out candidates is calculated. A method and an apparatus have been proposed in which a character extraction result is defined by a combination in which the evaluation function is defined and the evaluation function is optimized.

As another conventional method for detecting a character cut-out position, a projection pattern is obtained from an electronic image and a height of the projection pattern becomes a certain value or less, or a valley of the projection pattern. A method of detecting a cutout position by searching for a deepest position has been proposed. For example, literature (Information Processing Society of Japan, 33 volumes,
No. 9, page 1083, “Document processing system using high-precision character segmentation processing based on character structure information” (see “6. Separation processing of contact characters” in the middle of the right column on page 1085) Referenced.

[0004]

However, the above-mentioned conventional system has the following problems.

[0005] First, in the character extraction method described in Japanese Patent Application Laid-Open No. 6-203201, it is possible to extract a character composed of a plurality of parts, such as a biased and unstructured character. However, if they come into contact with each other, the contacted parts cannot be separated.

Further, in the method described in the above document, it is difficult to obtain an appropriate character cutout position candidate when character contact is not point-like but planar. Furthermore, when there are characters such as hyphens whose projection pattern is uniform and has little variation over a wide range, it is difficult to extract an appropriate cutout position candidate, or a redundant character cutout position candidate is extracted. Has the problem that

SUMMARY OF THE INVENTION Accordingly, the present invention has been made in view of the above problems, and an object of the present invention is to provide a character extracting method and a character extracting method capable of appropriately extracting a contacted character.

Another object of the present invention is to provide a character extracting method and method that does not perform excessive extracting.

It is another object of the present invention to provide a character segmenting method which can be easily realized by hardware.

[0010]

[Means for Solving the Problems]

Further, the character extracting method of the present invention comprises:
(A) Dividing an optically captured and digitized character line image into pixel blocks to be connected (referred to as “connected pixel blocks”),
(B) detecting the contour of the connected pixel block to determine the number of contours and the contour coordinates; (c) determining the determined number of contours;
The presence / absence of character contact is determined by comparing the inside / outside type of the outline and the position of the outline with a predetermined character contact determination condition. (D) When it is determined that there is a contact The pixel is tracked and scanned in the character string direction with a predetermined distance as an upper limit from the end point of the inner contour line, and an intersection position with the outer contour line or a position separated by a predetermined distance is determined. (E) setting a character end point and setting a character cutout boundary in the vertical direction with respect to the set character end point and the arrangement of the character lines. And

[Outline of the Invention] The outline of the present invention will be described below. In the present invention, means for determining a contact for performing contact determination and character cutout position estimation (contour tracking section 3, contour type discriminating section 4, contour end point detecting section 5), means for determining the presence or absence of contact (character contact) The determining unit 6) includes means for searching for a character cutout position when it is determined that there is a contact (character end detecting unit 7), and means for setting a character boundary (character cutout position setting unit 8). In the present invention, for the cutout position candidate, only a position uniquely estimated from the position of the inner contour is selected, so that there is an operational effect that redundant candidates do not occur. Further, according to the present invention, by performing the separation of the characters after the determination of the contact, there is an operational effect that unnecessary cutout processing can be avoided.

[0013]

Embodiments of the present invention will be described in detail with reference to the drawings. FIG. 1 is a block diagram showing a configuration of an embodiment of the present invention. Referring to FIG. 1, in a preferred embodiment of the present invention, (a) an image storage unit 1 for storing a digitized character line image captured by means such as optical scanning, and (b)
Image dividing unit 2 for separating a character line image into individual connected pixel blocks
(C) a contour tracing unit 3 that performs contour tracing processing on the connected pixel soul to detect a contour, and (d) discriminates an inner contour and an outer contour of the contour obtained by the contour tracing unit. A contour type discriminating unit 4, (e) a contour end point detecting unit 5 for detecting an end point of the contour line, and (f) a judgment of the presence / absence of contact based on the number of contour lines, a contour end point position, and a contour line type. , Character contact determination unit 6
(G) a character end detecting unit 7 for detecting a character end from the coordinates of a contour end point; and (h) a character cutout boundary setting unit 8 for setting a character cutout boundary at a position of a character end in a direction perpendicular to a character line. And (i) an overall control unit 9 for controlling the operation of each module.

The image storage unit 1 is fetched by means such as optical scanning and is stored in an OCR (Optical C).
A character line image cut out using a known character line cutout technique generally used in a character reader (optical character reading) device or the like is stored in an electronic form.

The image division unit 2 reads the digitized character line image stored in the image storage unit 1 and divides it into individual connected pixel blocks by a method such as labeling.

The contour tracing unit 3 performs a contour tracing process on the divided connected pixel blocks to obtain information such as the number of contour lines, a tracing direction, and circumscribed rectangle coordinate values of the contour lines.

The contour type discriminating unit 4 determines whether the contour is an inner contour or an outer contour from the tracking direction of the contour information obtained by the contour tracing.

The contour end point detector 5 detects the end points of the contour line obtained by the contour tracing.

The character contact judging section 6 judges the presence or absence of character contact with reference to the number of contour lines and the coordinates of the end points of the inner contour line and the outer contour line.

The character end point detecting section 7 tracks pixels along the character line direction from the end point of the inner contour line, and sets an intersection position with the outer contour line or a position separated by a predetermined distance as a character end point. I do.

The character cutout boundary setting section 8 sets a vertical line passing outside the set character end point as a character cutout boundary.

The overall control unit 9 controls the operation of the entire module including the control for extracting characters from each connected pixel block generated by the pixel dividing unit 2.

FIG. 2 is a flowchart for explaining the processing operation of the embodiment of the present invention. The operation of the embodiment of the present invention will be described with reference to FIGS. The control of each unit (module) is performed by the overall control unit 9.

The character line image stored in the image storage unit 1 is
The image is divided by the image dividing unit 2 into individual connected pixel blocks (step 102).

In order to obtain information for use in character contact determination for each of the divided connected pixel blocks,
Contour tracing processing is performed using the contour tracing unit 3 to obtain contour line information (step 104).

Further, using the contour type discriminating section 4 and the contour end point detecting section 5, the inside / outside type of the contour line is determined, and the end point of each contour line is detected (step 105).

Referring to the information on the contour line obtained in this way, a contact determination process is performed on the connected pixel block (step 106). Detection (step 107) and setting of character boundaries (step 108) are performed.

If it is confirmed that the above-described contact determination processing has been performed on all the connected pixel blocks (step 103)
No branch), and the process ends (step 109). Each of the above steps can be realized by executing a program in the information processing device, or can be realized by a dedicated hardware device or a cooperation of software and a hardware device.

[0029]

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a block diagram showing an embodiment of the present invention;

Referring to FIG. 1, character contact determination section 6 determines whether or not a character in the connected pixel block has touched. In this determination, the number of contour lines, the type of contour line (inside / outside), and the position of the contour line are referred to as arguments, and the presence / absence of contact is determined based on the following determination formula.

Assuming that the number of touching characters is two, the following equation (1) can be used as the equation.

If (the number of inner contours = 1), (the number of outer contours = 1) and (the inner contour is contained in the left half or the right half of the outer contour), it is determined that "there is a contact". Otherwise, it is determined that there is no contact. … (1)

Other than the above, the following equation (2)
Can be used.

｛(The number of inner contours = 1) or (the number of inner contours =
2)｝ and (the number of outer contours = 1) and ｛(when the number of inner contours = 1, the inner contour falls within the left half or right half of the outer contour) or (when the number of inner contours = 2, the inner contour Horizontal direction
If the overlap of the directions is within a predetermined ratio), it is determined that “there is contact”, and otherwise, it is determined that there is no contact. … (2)

Further, a determination formula when assuming that the number of contact characters is three or more can be defined as the following formula (3).

(The number of inner contours = 1), (the number of outer contours = 1) and (the distance between the left end of the inner contour and the left end of the outer contour is within 1/3 of the width of the outer contour) and If (the distance between the right end of the inner contour line and the right end of the outer contour line is within 1/3 of the width of the outer contour line), it is determined that "there is a contact", otherwise, "there is no contact" Is determined. … (3)

It is also possible to define a determination formula for determining the contact only with the inner contour line without assuming the number of contact characters. For example, the following determination formula (4) can be used. In this case, there is no need to consider the number of characters in contact.

If (the number of inner contours ≧ 1), “there is contact”
Otherwise, it is determined that there is no contact. …
(Four)

A judgment formula obtained by appropriately combining a plurality of these judgment formulas is used in the same manner.

When it is determined that there is a character touch, the character end point is detected by the character end detecting unit 7. Hereinafter, the case of the left end will be described as an example.

With reference to the left end point of the inner contour obtained by the contour end point detector 5, pixels are searched toward the left side, and a position intersecting with the outer contour is obtained. If an intersection position is found within the distance specified separately, that position is defined as a character end point. Similarly, the position of the character end point is defined for the right end.

Here, the distance separately specified in the character edge detecting section 7 is a value set with reference to the thickness of the character. In the case of handwritten characters, the setting is made with reference to writing implements, and in the case of printed characters, information relating to print fonts and character sizes. More specifically, when targeting a writing instrument such as a felt-tip pen, for example, it is preferably 2 mm.
The number of pixels is set to about the same length, and when a document printed by a word processor or the like is targeted, the length is preferably set to about 0.5 mm. Here, the expression of mm (millimeter) is used because the numerical value when converted into the number of pixels changes according to the resolution of the captured image.

The present embodiment will be described in further detail with reference to a specific example. In the following, a specific input character string pattern is assumed, taking the case of using the decision formula (2) as an example, and the operation of the present embodiment. Will be described.

For example, "120" as shown in FIG.
The case where the contact character is input will be described as an example. FIG. 3B is a partially enlarged view of the character pattern of FIG. 3A, wherein 301 is an inner contour, 302 is an outer contour, 303 is the left end of the inner contour, and 304 is
Indicates the left end of the character, 305 indicates the right end of the inner contour line, 306 indicates the right end of the character, and 307 indicates the character boundary.

The character pattern "120" is first divided by the image dividing section 2 into two pixel blocks "1" and "20" (step 102 in FIG. 2). Next, a contour tracking process and a process for finding a contour end point are performed on the pattern “1” (steps 104 and 105 in FIG. 2).

In the pattern of “1”, the number of outer contours = 1 and the inner contour = 0. In this case, since the character contact condition of Expression (2) is not met, the character contact determination unit 6 determines that there is no contact character, does not separate the character, and proceeds to the processing of the next pattern (step 106). “No” branch).

Next, contour information is checked for the portion "20". In this case, the outer contour = 1, the inner contour = 1, and the inner contour of “0” is located substantially in the right half of “20”.
Since this satisfies the character contact determination condition shown in the above equation (2), it is determined that there is a character contact ("Yes" branch in step 106).

An inner contour 301 of "0" is detected from "20", black pixels are tracked from the left end to the left side, and a point intersecting the outer contour 302 is obtained (step 107). This position corresponds to the left end 304 of the character (“0”). The cutout boundary 307 is set perpendicular to the character string direction along the left end of the character (step 108). The cutout position can be similarly obtained for the right end (step 1).
07, 108). However, in this case, a boundary for character extraction is set, but there is no pixel on the right side of the boundary, so that the extraction processing need not be performed as a result.

As another example, when a contact pattern “140” as shown in FIG.
The operation of the present embodiment will be described. First, “1” and “4”
It is the same as in the case of the above-mentioned “120”, and the same is true of the contact determination result for the part of “1.” Note that FIG. 4B shows the character pattern of FIG. In FIG. 4B, 401 is an inner contour, 402 is an outer contour, 403 is an inner contour left end, 404 is a character left end, and 405 is an inner contour right end. , 406 indicate the right end of the character, and 407 indicates the character boundary.

Next, contour information for the portion "40" is examined. In this case, if the top of “4” is closed,
There will be two inner contours. The adjacency relationship of the inner contour is determined, and it is determined that there is a contact because the overlap is within a predetermined value. If the upper part of “4” is far away, it is the same as the case of “120” described above.

Tracking is performed to the left from the left end of the detected inner contour of “0” to find an intersection with the outer contour. In this case,
No intersection with the outer contour is found within a predetermined distance. In this case, the character separation position is set at a predetermined distance separately determined.

Next, the case where the above-mentioned judgment formula (3) is used will be described with reference to an example. The operation when the contact pattern “502” as shown in FIG. 5A is input will be described. In this example, since the image division unit 2 cannot divide the image into individual connected pixel blocks, the image division unit 2 passes the data to the next step as it is. 5B is a partially enlarged view of the character pattern of FIG. 5A. In FIG. 5B, reference numeral 501 denotes an inner contour, 502 denotes an outer contour, and 503 denotes a contour. Is the left end of the inner contour line, 504 is the left end of the character, 505
Is the right end of the inner contour line, 506 is the right end of the character, and 507 is the left boundary of the character.

When examining the contour information, the number of outer contours = 1 and the number of inner contours = 1. Since the inner contour is a part of the middle character, the left end and the right end of the inner contour satisfy the condition of the determination formula (3). Therefore, this pattern is determined to be a contact character and sent to the character cutout processing. In this example, the end point of the middle character is detected as a cutout position on both the left and right sides.

The contour tracing processing including the labeling processing, the contour type determination processing, and the contour end point coordinate detection processing can be performed by software on a general-purpose processor.
It can also be realized as dedicated hardware having an image memory and an arithmetic device. Further, the contour tracing process in the contour tracing unit 3 can be performed simultaneously as a partial process of the labeling process in the pixel dividing unit 2, or can be used as a function module.

In the above-described embodiment of the present invention, since only a position uniquely estimated from the position of the inner contour is selected for the cutout position candidate, no redundant candidate is generated.
This has the operational effect of:

In the above-described embodiment of the present invention, the character is separated after the determination of contact is made, so that an unnecessary cutout process can be avoided.

[0057]

As described above, according to the present invention,
The following effects are obtained.

The first effect of the present invention is that the character can be cut out for more contact patterns than in the conventional method because it is not affected by the contact form of the character.

The second effect of the present invention is that (a) redundant character extraction processing is not performed, and (b) hardware can be easily implemented, so that character extraction processing can be performed at high speed. That is,

[Brief description of the drawings]

FIG. 1 is a block diagram showing a configuration of an embodiment of the present invention.

FIG. 2 is a flowchart illustrating an operation of the exemplary embodiment of the present invention.

FIG. 3 is a diagram for explaining the operation of one embodiment of the present invention.

FIG. 4 is a diagram for explaining the operation of one embodiment of the present invention.

FIG. 5 is a diagram for explaining the operation of one embodiment of the present invention.

[Explanation of symbols]

DESCRIPTION OF SYMBOLS 1 Image storage part 2 Image division part 3 Outline tracking part 4 Outline type discrimination part 5 Outline end point detection part 6 Character contact judgment part 7 Character end detection part 8 Character cutout boundary setting part 9 Overall control part 101 Overall control part 11 stores an image. A step of fetching a character line image from the unit 1 and starting a character segmentation process 102 a step of the image dividing unit 2 dividing the character line image into individual connected pixel blocks 103 a process for all connected pixels by the overall control unit 11 Step 104 in which it is determined whether or not the processing has been performed. Step 104 for determining the outline information of the connected pixel block by the outline tracking section 3, outline type determination section 4, and inner outline left end. 105 Inner outline left end detection section 4 and inner outline right end detection section 5 determine the inner outline. Step 106 for detecting the left and right edges of the outer contour The step for detecting the left and right edges of the outer contour by the outer contour left edge detector 6 and the outer contour right edge detector 7 107 Step in which the character contact determination section 9 makes contact determination 108 Step in which the character cutout boundary setting section 10 sets a character cutout boundary 109 The overall control section 11 sends the character cutout result to the subsequent processing, and ends the character cutout processing. Step 301 Inner outline 302 Outer outline 303 Inner outline left end 304 Character left end 305 Inner outline right end 306 Character right end 307 Character boundary 401 Inner outline 402 Outer outline 403 Inner outline left end 404 Character left end 405 Inner outline right end 406 Right end of character 407 Character boundary 501 Inner outline 502 Outer outline 503 Inner outline left 504 Character left end 505 Inner outline right end 506 Character right end 507 Character left boundary 508 Character right boundary

Claims (1)

(57) [Claims] (A) dividing a character line image optically captured and digitized into a pixel block to be connected (hereinafter referred to as "connected pixel block"); and (b) detecting an outline of the connected pixel block. Determining the number of contours and the coordinates of the contours; and (c) determining a predetermined character contact determination condition by referring to the determined number of contours, the type of inside / outside of the contours, and the position of the contours. The presence or absence of a character contact is determined by comparison. (D) If it is determined that there is a character contact, pixels are tracked in a character string direction with a predetermined distance as an upper limit from an end point of the inner contour line. Scan, and the intersection with the outer contour,
Or a position separated by a predetermined distance,
(E) setting a character cutout boundary in a direction perpendicular to the set of character lines with respect to the set character end point, comprising the steps of (a) to (e) described above. Character extraction method.