JP2899383B2 - Character extraction device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION (Industrial Application Field) The present invention relates to a character recognition device and the like for recognizing a document.
The present invention relates to a character extracting device for extracting a character pattern from a character string image recorded on a medium such as paper.

(Prior Art) Conventionally, as a technique relating to a character recognition device for extracting a character pattern by a character extraction device and recognizing the extracted character pattern, there is one disclosed in, for example, JP-A-63-16391.

In general, in a character recognition device for a character type having no character frame or a handwritten document, a character extraction device is used to extract a character string image for one line into a character pattern for each character, and perform recognition. Do.

In the character extracting device described in the above document, characters are extracted based on a peripheral distribution obtained by projecting a character string image in a direction perpendicular to the character string direction (line direction).
That is, a peripheral distribution is detected from a character string image, a pattern (sub character pattern) of a continuous area where the peripheral distribution value is 1 or more is extracted, and the width of the sub character pattern and the distance between adjacent sub character patterns are determined. , The character pattern is determined by combining the sub-character patterns.

(Problems to be Solved by the Invention) However, in the device having the above-described configuration, in the case where a character having no character frame, a handwritten document, and the like, adjacent characters are in contact with each other or overlap each other. , A sub-character pattern containing multiple characters is extracted.
As a result, character patterns cannot be cut out correctly.
Therefore, for example, when character recognition is performed using the cut-out character pattern, a problem that recognition accuracy is lowered occurs, and it has been difficult to solve the problem.

The present invention solves the problem of the related art that a character pattern cannot be cut out correctly when a character string image comes into contact with another character pattern or an overlapping character pattern exists. A character extracting device is provided.

(Means for Solving the Problems) In order to solve the above problems, the present invention provides a character extracting apparatus for extracting a character pattern from character string image data obtained from a character string on an input medium, comprising: a search area setting unit; It has a route detection unit, a boundary detection unit, and a character pattern cutout unit.

Here, the search area setting unit performs a search with a search start point and a search end point for setting a boundary line between adjacent partial patterns in the character string image data, and a line segment connecting the search start point and the search end point as axes. A function of setting an area and updating the search start point and the search end point based on the route obtained by the route detection unit and the validity of the route obtained by the boundary detection unit. . The path detection unit weights the pixel density value of each coordinate on the path by a predetermined coefficient determined according to the direction of the path at the coordinates, in the path in the search area from the search start point to the search end point, The path is detected so that the accumulated value is minimized. The boundary detection unit determines the validity of the route based on the route detected by the route detection unit and the pixel density value of a region near the route, and determines that the route is adjacent when the route is recognized as valid. It is determined that it is a boundary line between the partial patterns to be performed. Further, the character pattern cutout section has a function of cutting out a character pattern from the character string image data based on a position of a boundary between the adjacent partial patterns.

The search area may have a convex polygonal shape that is symmetric about a line segment connecting the search start point and the search end point and has the search start point and the search end point as vertices.

(Operation) According to the present invention, since the character segmenting apparatus is configured as described above, when character string image data is input to the search area setting unit, the search area setting unit sets a search start point, a search end point, , A search area centered on a line segment connecting the search start point and the search end point, and the set value is given to the path detection unit, the boundary detection unit, and the character pattern cutout unit. The route detector detects an optimal route based on the input search area, and sends the detection result to the boundary detector. The boundary detection unit determines the validity of the path based on the path and the pixel density value of the neighboring area, and sets the path as a boundary between adjacent partial patterns when the path is recognized as valid. The setting result is given to the character pattern cutout section. Then, the character pattern cutout unit cuts out the character pattern from the character string image based on the position of the boundary between adjacent partial patterns.
As a result, when adjacent characters are in contact with each other,
Even when they overlap, each character pattern can be accurately extracted.

Also, when the search area is a convex polygon, accurate path detection can be performed with a small amount of calculation.

 Therefore, the above problem can be solved.

(Embodiment) FIG. 1 is a functional block diagram of a character extracting apparatus according to an embodiment of the present invention.

In FIG. 1, a character string image input unit 1 for reading a character string on an input medium and outputting character string image data is provided, and on the output side, a character pattern is cut from the character string image data one character at a time. An output character extracting device 10 is connected.

The character string image input unit 1 optically scans, for example, an information medium on which one or a plurality of character strings are written, and converts the scan result into a binary electric signal (image pattern) by photoelectric conversion. And further extracts a character string image included in the image pattern and extracts the character string image data into a character extracting device.
It has the function of outputting to 10. To extract a character string image from an image pattern, various character extraction methods such as a histogram method can be used.

The character cutout device 10 determines a search start point for setting a boundary between adjacent partial patterns in the character string image data input from the character string image input unit 1, a search end point, and a search start point and a search end point. A search area setting unit 11 is provided for setting a search area around a connected line segment. Here, the partial pattern is, for example, a block of individual black pixels in the character string image. The output of the search area setting unit 11 includes a route detection unit 1
2. The boundary detection unit 13 and the character pattern cutout unit 14 are connected.

The path detection unit 12 weights the pixel density value of each coordinate on the path with a predetermined coefficient determined according to the direction of the path at the coordinates in the path in the search area from the search start point to the search end point. It has a function of detecting the route so that the accumulated value is minimized. The route detection unit 12 includes a route direction data calculation unit 12a that calculates route direction data corresponding to each pixel in the search region based on the pixel density value in the search region, and a route that calculates a route based on the route direction data. Calculation unit 12b
The boundary detection unit 13 is connected to the output side.

The boundary detection unit 13 determines the validity of the path based on the detected path and the pixel density value of the neighboring area. It has a function to determine that it is a line, and its output side has
The character pattern cutout unit 14 is connected. The character pattern cutout unit 14 cuts out a character pattern from character string image data based on the position of a boundary between adjacent partial patterns.

FIG. 2 is an operation flowchart of FIG. 1. Referring to FIG. 2, a character extracting operation from input of character string image data from the character string image input unit 1 to output of a character pattern OUT will be described. (1) to (6) will be described below.

(1) Search Area Setting (Step S1) In step S1, when the search area setting unit 11 detects any of the following three states (a1) to (a3), the search area setting unit 11 sets the search area according to each case. And outputs the corresponding image data to the path detection unit 103, and proceeds to step S2.

(A1) When the character string image data is input from the character string image input unit 1 The search area setting unit 11 stores the character string image data and sets a search area at the head of the pattern of the character string image data.

(A2) When no valid boundary line is detected by the boundary line detection unit 13, an area obtained by shifting the already set search area by a predetermined value in the character string direction (for example, the row direction) is replaced with a new search area. Set as

 The shift value is, for example, 1/2 of the search area width.

(A3) When a valid boundary is detected by the boundary detection unit 13 A new search area is set based on the position of the boundary. For example, the search area is set by aligning the leading end of the search area with the rearmost position of the coordinates forming the boundary on the pattern composed of the character string image data.

FIG. 3 is a diagram showing an example of a search area. In the figure, an area 20 surrounded by a dotted line represents an image pattern composed of character string image data, and a hatched rectangle represents a black pixel included in an image pattern composed of character string image data. For example, if the row direction is represented by X and the direction perpendicular to the row direction is represented by Y, the search area 21 is a pixel above the top of the character line by A pixels.
21a is set as a search start point, and a pixel which is below the lower end of the character line by A pixels and whose X coordinate is equal to the pixel 21a, that is, a pixel 21b is set as a search end point, and a line connecting the search end point 21a and the search end point 21b is set as an axis. As a hexagonal area having a width of 2A + 1.

(2) Calculation of route direction evaluation value data g (X, Y) and route direction data d (X, Y) (Step S2) In Step S2, the route direction data calculation unit 12a
Input the image data of the search area from the search area setting unit 11,
Path direction data d (X, Y) corresponding to the pixel density value f (X, Y) of the pixel (X, Y) in the search area, and a path for calculating the path direction data d (X, Y) Direction evaluation value data g
(X, Y) are generated by the following two steps S2-1 and S2-2.

Here, for example, the pixel density value f (X, Y) is set to 10 for a white pixel (background portion) and 100 for a black pixel (character portion). Also the coordinates of the search start point and search end points and (X _O, Y _S) and (X _O, Y _E).

(2) (i) Step S2-1 path direction in the search start point data d (X _O, Y _S) to set the initial value and the path direction evaluation value data g (X _O, Y _S) on.

_{d (X O, Y S)} = 1 ...... (1) g (X O, Y S) = 0 ...... (2) (2) (ii) Step S2-2 X-direction in the main scanning direction, a Y-direction Sub-scanning direction, pixel (X _O
-1, Y _S +1) scan start position, the pixel (X _O, scans the search area 21 Y _E) as a scanning end position, each pixel (X, Y) path direction data d (X each, Y ) And the route direction evaluation value data g (X, Y) are calculated by the following equation (3).

g (X, Y) = min (G ₀ , G ₁ , G ₂ ) (3) Where G ₀ = g (X−1, Y−1) + K ₀ f (X, Y) G ₁ = g (X, Y−1) + K ₁ f (X, Y) G ₂ = g (X + 1, Y _{-1) + K 2 f (X} , Y) ...... (5) If g (X−1, Y−1) is outside the search area 21 in equation (5), g (X−1, Y−1) = ∞.
(Infinity). The same applies to g (X, Y-1) and g (X + 1, Y-1) in the expression (5).

FIG. 4 shows three pixels P ₀ (X−X) adjacent to the pixel P (X, Y).
1, Y-1), P 1 (X, Y-1), is a diagram illustrating P ₂ a (X + 1, Y-1 ). The route direction data calculation unit 12a will be described in more detail with reference to FIG.

Equation (5) shows that three pixels P ₀ ,
When assuming a path through the P _1, each of the pixels P of P _2, the in the adjacent pixels P _0, P _1, each of the P ₂ of path direction evaluation data, the route pixel density value of the pixel P It indicates a value obtained by adding values weighted by the coefficients K ₀ , K ₁ , and K ₂ determined according to the directions. Therefore, g represented by the equation (3)
(X, Y) is an arbitrary path from the pixel (X _O , Y _S ) to the pixel (X, Y), and determines the pixel density value of each pixel on the path according to the direction of the path at the pixel. This is the minimum cumulative value in the case where the cumulative value of the values weighted by the predetermined coefficients defined in the above becomes the minimum.

Further, d (X, Y) expressed by the equation (4) is a pixel (X, Y)
At the time of the previous scanning line (Y
-1) is the data indicating the path direction from pixel P ₀ ,
When the minimum cumulative value is obtained on the route passing through P ₁ and P ₂ , values 0, 1, and 2 are obtained.

5 and 6 show the route direction evaluation value data g (X, Y) and the route direction data d for the search area 21 in FIG.
It is a figure showing an example of (X, Y).

When the process in step S2 ends, the process proceeds to step S3.

(3) Path Detection (Step S3) In step S3, the path calculation unit 12b sends the path direction data d corresponding to each pixel from the path direction data calculation unit 12a.
(X, Y) by entering the pixel (X _O, Y _S) from the pixel (X _O, Y
_E ) Among the arbitrary routes to the route, the route in which the cumulative value of the values obtained by weighting the pixel density values of the respective pixels on the route with a predetermined coefficient determined according to the direction of the route at the pixel is the minimum. To
It is determined by the following steps S3-1 to S3-3, and the position data of the pixel group forming the path is output to the boundary detection unit 13.

(3) (i) Step S3-1: Y = Y _E , and X coordinate DX of the path at the sub-scanning position Y
(Y) a DX (Y) = and X _O.

(3) (ii) Step S3-2 Subtract Y by 1, and if d (DX (Y + 1), Y + 1) is 0, DX (Y) = DX (Y + 1) -1 d (DX (Y + 1) , Y + 1) is 1, if DX (Y) = DX (Y + 1) d (DX (Y + 1), Y + 1) is 2, then DX (Y) = DX (Y + 1) +1 until Y becomes equal to Y _S This step S3-2 is repeated.

(3) (iii) Step S3-3 coordinates (DX (Y), Y) pairs of _{(Y = Y S ~Y E)} , determined as the path.

As described above, d (X, Y) is the pixel density value of each pixel on the path from the pixel (X _O , Y _S ) to the pixel (X, Y). Is the data indicating the path direction from the previous scanning line (Y-1) when the cumulative value of the values weighted by the predetermined coefficients determined according to the direction of the path becomes the minimum, and therefore, is determined in steps S3-1 to S3-1. based on S3-3, by determining the coordinates corresponding to the path direction while subtracting one sequentially from Y Y = Y _E, optimum leading to _{(X O, Y E) ~} (X O, Y S) The route can be determined.

In FIG. 6, the pixels with circles represent the steps in this step.
The coordinates on the route obtained by S3-3 are shown. Also, the seventh
The figure is a diagram in which the result of step S3-3 for the route direction data d (X, Y) in FIG. 6 is associated with the character string image pattern in FIG. 3, and the pixels indicated by black circles in the figure are detected. Represents the pixel on the path.

When step S3 ends, the process proceeds to steps S4 to S6.

(4) Boundary Line Detection (Steps S4 to S6) In step S4, the boundary line detection unit 13 sets the route calculation unit
Input the position data of the pixel group forming the path from 12b,
Further, the image density value of a region near the route is read from the search region setting unit 11, and the validity as a boundary between adjacent partial patterns of the route is determined.

In this determination, (b1) the length of the boundary between adjacent partial patterns is short, and (b2) the number of black pixels on the boundary is small.
(B3) This is performed based on a general tendency regarding a boundary line such as a large number of black pixels included between adjacent boundary lines. For example, when all of the following three conditions are satisfied, the route calculation unit
The route obtained from 12b is determined to be valid.

Condition 1 L <K ₁ × H; corresponding to the property of the above (b1)... (7) Condition 2 B ₁ <K ₂ × H × LW; corresponding to the property of the above (b2)... (8) Condition 3 B ₂ > K ₃ × H × LW; corresponding to the property of (b3) above (9) where L, B ₁ , and B ₂ are the path length, the number of black pixels on the path, and the distance from the adjacent boundary line, respectively. This is the number of black pixels included between them.
H is the size (height) in the direction perpendicular to the row direction of the character string image data. LW is the average line width of the character pattern included in the character string image data. The average line width can be detected in various ways. K ₁ , K ₂ , and K ₃ are predetermined constants determined empirically. For example, values of K ₁ = 1.5, K ₂ = 0.1, and K ₃ = 0.4 are set.

In the initial operation of the character extraction device 10,
It is assumed that a boundary line is virtually set at the head position of the character string image data.

If the validity of the route is not recognized in step S4, the boundary detection unit is set to set the next search area.
13 sends an instruction to the search area setting unit 11, and returns to step S1.

When the validity of the path is confirmed, in step S5, the boundary detection unit 13 determines the path as a boundary between adjacent partial patterns, and determines position data of a pixel group forming the boundary. save. Then, in step S6, the boundary detection unit 13 determines whether to end the boundary detection processing. If the detection process is not completed, the search area setting unit 11 executes step S1 to set the next search area. On the other hand, when ending the detection processing, the character pattern cutout unit 14 executes step S7.

In determining whether or not to end the boundary line detection processing, it is determined whether or not black pixels do not exist in an area behind the boundary line in the character string image data. It is determined that the processing is not to be terminated, and if not, it is determined that the detection processing is to be terminated.

(5) Generalization from Search Area Setting to Boundary Line Detection (Steps S1 to S6) FIG. 8 is a diagram showing an example of detecting a boundary line between partial patterns in steps S1 to S6. In the figure, reference numeral 30 denotes an example of a pattern composed of character string image data, in which the row direction is indicated by Y and the direction perpendicular to the X row direction.

When the search area setting unit 11 inputs the character string image pattern 30, the search area 31 is set, and the path detection unit 12 detects a path indicated by a dotted line. The boundary detection unit 13 detects that the condition of the expression (9) is not satisfied, regards the route as invalid, and then sets the search area 32 by the search area setting unit 11. For the search area 32, the route detection unit 12 detects the route indicated by the dotted line, but the boundary detection unit 13 detects that the condition of the expression (8) is not satisfied, and considers the route to be invalid. ,
Next, the search area setting unit 11 sets the search area 33.

Regarding the search area 33, the path detection unit 12 detects the path indicated by the dotted line, the boundary line detection unit 13 determines that the path is valid, stores the position data of the path, and then the search area setting unit 11 The search area 34 is set. Thereafter, by repeating the same processing, four sets of boundary lines indicated by dotted lines are obtained as shown in the area 35. The boundary line detection unit 13 finally stores the position information of the four sets of boundary lines. As a result, substantially four partial patterns P1 to P4 are obtained.

(6) Character Pattern Extraction (Step S7) In step S7, the character pattern extraction unit 14 reads out the position data of the boundary between the partial patterns from the boundary detection unit 13, and sets the search area based on the position of the boundary. From the character string image data stored in the unit 11, a character pattern for each character is cut out, the data of the corresponding character pattern OUT is output, and the operation of the character cutout device 10 ends.

That is, in the character cutout processing, it is necessary to determine the cutout position based on the positions of the individual partial patterns constituting the character pattern and the pattern data. Therefore, in the character extraction method according to the present embodiment, a partial pattern is extracted based on the position of the boundary line, and each of the partial pattern and an integrated pattern obtained by combining the partial patterns is used as a character candidate pattern. Detect geometric character evaluation values. Then, based on the geometric character evaluation value, an optimum character candidate pattern arrangement is obtained and used as a character cutout result.

Here, when setting the integrated pattern, the integrated pattern width is set to a predetermined threshold (for example, 1.5 times the height of the character string image pattern).
Make settings within a range that does not exceed (double). Character candidate pattern i
As the geometric character evaluation value Vi for, for example, the following expression (10) is used, and a smaller value indicates a more likely character.

However, Wi: the pattern width of the character candidate pattern i, H: the character string image pattern height In detecting the optimal character candidate pattern arrangement, the character candidate pattern arrangement represented by the following equation (11) is used. Overall evaluation value for Is the smallest, the character candidate pattern arrangement is taken as the character extraction result.

Note that ni: the number of partial patterns constituting the character candidate pattern i. FIG. 9 is a diagram showing an example of a geometric character evaluation value for the example of boundary detection (the area 35) in FIG. Based on this figure, the character candidate pattern arrangement when equation (11) is minimized is obtained as follows: "East" (P1), "Kyo" (P2), "To" (P3 +
P4) is obtained as a character segmentation result.

Therefore, this embodiment has the following advantages.

(I) In the route in the search area from the search start point to the search end point, the route detection unit 12 determines a pixel density value of each coordinate on the route by a predetermined coefficient determined according to the direction of the route at the coordinates. Weighting is performed, and a path for detecting a boundary line is detected so that the accumulated value is minimized. Therefore, the following two properties (i
-1) and (i-2) can be obtained.

(I-1) The ratio of character-constituting pixels (pixels with a high pixel density value) crossing the path is small.

(I-2) The deviation of the route from the line segment connecting the search start point and the search end point is small.

As described above, in the present embodiment, when adjacent characters of the input character string image overlap each other, a path having a small deviation from a line segment connecting the search start point and the search end point from the background area (white pixel area) between the adjacent characters is determined. Is determined as a boundary line between adjacent partial patterns, so that the character pattern OUT of each adjacent character can be correctly extracted. Further, when adjacent characters are in contact with each other, a path having a small deviation from a line segment connecting the search start point and the search end point and having a small passage distance within a block of character constituent pixels including the contact point is Since it is determined as a boundary between adjacent partial patterns, each character pattern OUT of adjacent characters can be extracted with high accuracy. As a result, the accuracy of character extraction is improved as compared with the related art, so that it is possible to avoid a reduction in the recognition rate due to, for example, a character extraction error in the character recognition device, and to realize a highly accurate character recognition device.

Note that the present invention is not limited to the above embodiment, and various modifications are possible. For example, there are the following modifications.

(1) In order to calculate the route direction evaluation value data g (X, Y), each of the parameters K ₀ , K ₁ , and K ₂ is determined as in equation (6). K ₀ , K ₁ ,
It is preferable to set the K _2.

For example, in the case of a character string image (for example, a character string of a print type) composed of characters having no significant overlap or contact, K ₁
If K ₀ and K ₂ are set to be larger than K ₁ and larger than K ₁ , the route is determined from an area where the deviation from the line segment connecting the search start point and the search end point is small.

(2) In FIG. 3, the position and shape of the search area 21 are set to the width 2A.
Although a hexagon having +1 is set, the position and shape of the search area 21 may be changed as appropriate.

For example, a narrower search area is set than in the case of a character string image (for example, a character string of printed characters) composed of characters having no significant overlap or contact, and conversely, significant overlap or contact exists. In the case of a character string image composed of characters (for example, a handwritten character string), a wider search area may be set. Furthermore, the shape of the search area 21 is changed to a rectangle other than a hexagon, an octagon,
It may be elliptical or the like.

In the above embodiment, the search start point and the search end point are set outside the area of the character string image. However, the search start point and the search end point may be set within the area of the character string image as needed.

(3) Three pixels (X−1, Y−1) and (X−1, Y−1) with respect to the pixel of interest (X, Y) for calculating the route direction evaluation value data g (X, Y)
Route direction evaluation value data g (X-1, Y-1), g (X, Y-1), g (X + 1, Y-) at (X, Y-1), (X + 1, Y-1)
Although the character extracting device 10 is configured to refer to 1), the number of pixels to be referenced may be increased or decreased as necessary. For example,
To calculate g (X, Y), five pixels (X−2, Y−1),
(X-1, Y-1), (X, Y-1), (X + 1, Y-1),
Referring to (X + 2, Y-1), it is possible to correctly detect the boundary between the partial patterns even for the overlapping characters and the touching characters that are more involved than in the above embodiment.

(4) Although the X coordinate of the search start point and the X coordinate of the search end point are set to be equal, it is preferable to appropriately change the X coordinate according to the characteristics of the target character string image. For example, if each character constituting the character string is obliquely inclined (for example, in the case of an English character string described in italics), the search start point and the search end point are set with respect to the direction perpendicular to the character arrangement of the character line image. It is preferable to set a diagonal straight line, set an appropriate search area around a line segment connecting the search start point and the search end point, and perform a process for extracting a character pattern. In this case, the parameters K ₀ , K ₁ ,..., K for calculating the route direction evaluation value data g (X, Y)
_m-1 (where m is the number of reference pixels for calculating g (X, Y))
May be set to a smaller value of the parameter corresponding to a direction closer to the direction of the diagonal straight line.

(5) Although the pixel density value corresponding to the white pixel is set to 10 and the pixel density value corresponding to the black pixel is set to 100, these numerical values may be changed as appropriate.

(6) The case where the input character string image is binary black and white data has been described. However, the present invention can be applied to the case where the input character string image is multivalued data.

(7) The case where the path is determined based on the pixel density value of the input character string image has been described. However, the path is determined based on the pixel density value of the output image obtained by performing some processing on the input character string image. Is also good. For example, the input character string image is divided into blocks of M × N pixels (where M and N are positive integers) determined based on the height of the character string image, the average line width, and the like. It is also possible to create a binary or multi-valued reduced pattern in which each divided block is defined as one pixel, and determine a path based on the pixel density value of the reduced pattern. At this time, if the input character string image is locally blurred or the pixel density value is high, a reduced pattern that eliminates these effects can be created, so that erroneous determination of the path can be avoided.

(8) Each block in the character extracting device 10 shown in FIG. 1 may be configured by an individual circuit using an integrated circuit or the like, or may be configured to be executed by program control using a microcomputer or the like.

(Effect of the Invention) As described in detail above, according to the present invention, the search area setting unit sets the search area of the character string image, and the path detection unit detects the optimal path based on the search area, The boundary detection unit sets a boundary between the partial patterns based on the route and the pixel density value of the neighboring area, and further, the character pattern cutout unit cuts out the character pattern based on the boundary. . Therefore, the following 2
A path having two properties (i) and (ii) can be obtained.

(I) The ratio of character-constituting pixels (pixels having a high pixel density value) crossing the path is small.

(Ii) The deviation from the line connecting the route search start point and the search end point is small.

Accordingly, when adjacent characters of the input character string image overlap with each other, a path having a small deviation from a line connecting the search start point and the search end point from the background area between the adjacent characters is set as a boundary line between adjacent partial patterns. Therefore, the character pattern of each of the adjacent characters can be correctly extracted. Further, when adjacent characters are in contact with each other, a path having a small deviation from a line segment connecting the search start point and the search end point and having a small passage distance within a block of character constituent pixels including the contact point is Since it is determined as a boundary line between adjacent partial patterns, each character pattern of adjacent characters can be extracted with high accuracy.

Therefore, since the accuracy of character extraction is improved as compared with the conventional art, it is possible to avoid a decrease in the recognition rate due to a character extraction error in a character recognition device, for example, and to realize a highly accurate character recognition device.

In addition, when the search area is formed in a convex polygonal shape, the path can be detected with high accuracy and with a small amount of calculation.

[Brief description of the drawings]

FIG. 1 is a functional block diagram of a character extracting device showing an embodiment of the present invention, FIG. 2 is an operation flowchart of FIG. 1, and FIG.
FIG. 4 shows an example of a search area, FIG. 4 shows three pixels adjacent to a pixel (X, Y), FIG. 5 shows an example of route direction evaluation value data g (X, Y), FIG. 6 shows the route direction data d.
FIG. 7 is a diagram showing an example of (X, Y), FIG. 7 is a diagram showing an example of route detection by the route detection unit 12, FIG. 8 is a diagram showing an example of boundary line detection, and FIG. It is a figure showing an example of a value. 1 ... Character string image input unit, 10 ... Character cutout device, 11 ...
... Search area setting unit, 12... Route detection unit, 12a... Route direction data calculation unit, 12b... Route calculation unit, 13... Boundary line detection unit, 14.

────────────────────────────────────────────────── ─── Continuation of the front page (56) References JP-A-3-92987 (JP, A) JP-A-3-141482 (JP, A) JP-A-4-68481 (JP, A) (58) Field (Int.Cl. ⁶ , DB name) G06K 9/34

Claims (2)

(57) [Claims] A search start point and a search end point for inputting character string image data obtained from a character string on an input medium, and setting a boundary between adjacent partial patterns in the character string image data; A search area setting unit that sets a search area around a line segment connecting the start point and the search end point; and a pixel density value of each coordinate on the path in the search area from the search start point to the search end point. Is weighted by a predetermined coefficient determined according to the direction of the route at the coordinates, and a route detection unit that detects the route so that the accumulated value is minimized; a route detected by the route detection unit and the route A boundary detection unit that determines the validity of the path based on the pixel density value of the neighboring area, and determines that the path is a boundary between adjacent partial patterns when the path is recognized as valid. Boundary line between the adjacent partial patterns A character pattern cutout unit that cuts out a character pattern from the character string image data based on the position, wherein the search area setting unit is obtained by the path obtained by the path detection unit and the boundary line detection unit. A character extracting device, wherein the search start point and the search end point are updated based on the validity of the route. 2. The character segmenting device according to claim 1, wherein the search area is a convex polygonal shape that is symmetric with respect to a line segment connecting the search start point and the search end point and whose vertices are the search start point and the search end point. Character extraction device.