JP2639489B2 - Image data extraction method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image data extracting method for extracting image parts from image raster data including image parts such as characters, symbols and figures.

[0002]

2. Description of the Related Art Usually, in the technical field of image processing,
It is common to extract specific image components such as characters, symbols, and graphics, or a group thereof (for example, a group of characters such as idioms) from an image represented by raster data. By comparing a part with, for example, a standard pattern prepared in advance, character recognition by pattern recognition has been performed. As a method of extracting such image parts, there is a method of semi-automatically performing an operation instructing a computer to specify a region including image parts on raster data. That is, the operator operates a digitizer, a mouse, and the like while viewing the display on which the image is displayed, and designates an area including the image component on the raster data of the image. The computer extracts the area instructed from the raster data.

[0003] In addition to the extraction of such image parts, the usability of the extracted image parts is enhanced by giving information on the direction indicating the direction of the image part on raster data. For example, when the extracted image component is used for pattern recognition, a standard pattern of a character inclined according to the direction can be selected from the information on the direction, and the accuracy of matching with the standard pattern can be improved. By the way, such information on the direction of the image component is input to the computer by a digitizer, a mouse, or a key operation before or after the operator specifies an area including the image component.

[0004]

However, in such a conventional technique, it is necessary for an operator to input both an area for extracting an image part and a direction of the image part to a computer. There was a problem of bad.

[0005] The image data extraction method of the present invention has been made in view of the above problems, and has as its object to improve operability when extracting an image component from raster data of an image including the image component. .

[0006]

Means for Solving the Problems In order to achieve such an object, the following structure is adopted as means for solving the above problems.

That is, the image data extracting method of the present invention, as shown in FIG. 1, is an image data extracting method for extracting an image component from raster data of an image including the image component.
An area specifying step S1 for specifying a rectangular area including the image component in the raster data at two points at both ends of one diagonal of the rectangle, and the rectangular area specified in the area specifying step S1 A line segment connecting the two points, which is used to extract the image part and indicate the rectangular area.
And an image component extraction / direction calculation step S2 for calculating the direction of the image component based on the inclination of the image component.

[0008] Incidentally, the rectangle here means a rectangle and a square formed by raster scanning. The above structure
In the method for extracting image data of
Corresponds to any one of the four predetermined directions.
Can be determined.

[0009]

According to the image data extracting method of the present invention constructed as described above, in the area specifying step S1, a rectangular area including image parts in raster data is specified at two points at both ends of one diagonal of the rectangle. Then, in the image part extraction / direction calculation step S2, an image part is extracted from the rectangular area specified in the area specifying step S1, and a line segment connecting the two points serving as the basis of the rectangular area instruction is obtained . The direction of the image component is calculated based on the inclination .

Also, the direction of the image component is determined in advance.
To determine which of the four directions
The following four modes are conceivable in the configuration that has been completed . That is, there are a mode in which a rectangle formed by setting two points at both ends of one diagonal line is vertically long, and a mode in which the rectangle is horizontally long. Further, for each mode, a line segment connecting the two points, that is, a rectangle There is a mode in which one of the diagonal lines goes down to the right and a mode in which it goes down to the left. A total of four modes can be considered. For these four modes, four directions of the image component, for example, the upright direction, the inverted direction, the rightward direction, and the leftward direction are previously associated, and two
When a point is specified, it works to calculate the direction of the image component based on the inclination of the line connecting the two points.

[0011]

DESCRIPTION OF THE PREFERRED EMBODIMENTS In order to further clarify the configuration and operation of the present invention described above, a preferred embodiment to which the image data extracting method of the present invention is applied will be described below.

FIG. 2 is a block diagram showing a schematic configuration of a pattern recognition apparatus to which an image data extracting method according to one embodiment of the present invention is applied. As shown in FIG. 1, the pattern recognition apparatus includes a processing unit 10 having a well-known CPU 1, a ROM 3, and a RAM 5, an image scanner 12 for reading an image on a document as raster data, a pattern recognition instruction and raster data. Interactive terminal 14 for display
And a magnetic disk device 16 as an external storage device for storing various data.

The processing unit 10 is configured as an arithmetic and logic operation circuit. The CPU 1 that performs processing according to a program stored in the ROM 3 has an image input port 20 directly connected to the image scanner 12 via a common bus 17, It is connected to a terminal control unit 22 directly connected to the terminal 14 and a disk control unit 24 directly connected to the magnetic disk device 16.

The image scanner 12 optically scans an image on a document 12a placed on a drum, reads image data as binary black and white information, and expresses and outputs the binary information in a run-length format. Things. Manuscript 1
The image on 2a is a line drawing including image parts such as characters, symbols, and figures. The interactive terminal 14 includes a keyboard (not shown) for instructing various processes, a mouse 14a, and the like, and a CRT display 14b for displaying the results of the processes. The magnetic disk device 16 stores in advance a dictionary 16a in which standard patterns for recognizing image parts such as characters and figures are collected.

The CPU 1 stores raster data of an image read by the image scanner 12 in the magnetic disk device 16 in accordance with an instruction from the interactive terminal 14, and extracts image parts such as characters, symbols, and figures from the raster data. Then, the image part is collated with the standard pattern of the dictionary 16a stored in the magnetic disk device 16, and the collation result is output to the magnetic disk device 16 or the CRT display 14b of the interactive terminal 14. Interactive terminal 14
Performs such a series of pattern recognition operations in the processing unit 1.
0, and an area point required for the pattern recognition work is input.

Next, details of the pattern recognition work performed by the processing unit 10 will be described with reference to the flowcharts of FIGS. When the processing unit 10 starts processing, it first performs processing for inputting an image (step 100). This image input is performed by taking in the run-length-rendered raster data read from the image scanner 12 through the image input port 20 and developing the raster data in a predetermined area of the RAM 5. Next, the image data is stored in a predetermined area of the magnetic disk device 16, and is output to and displayed on a CRT display (steps 200 and 300).

Subsequently, an image part extraction process for extracting image parts such as characters, symbols and figures from the raster data developed in the RAM 5 in step 100 is performed (step 40).
0). In the image part extraction process, as shown in FIG. 4, first, two area designation points P1 and P2 input from the mouse 14a are read (step 410). here,
How the mouse 14a is used to specify the area designated points P1, P2
Will be described. The operator looks at the CRT display 14b on which the image is displayed, and
A is operated to input the area designating points P1 and P2. More specifically, as shown in FIG. 6, a rectangle (a diagonal line connecting two points) is displayed on the screen G of the CRT display 14b. When an area (shown by a broken line) is formed, two points, that is, two area designating points P1 and P2 are input so that an image part such as the character "A" is included in the rectangular area.

Subsequently, the coordinates (x1, y1), (x1) are used to determine to which position the area designated points P1, P2 read in step 410 correspond on the image data, ie, the raster data.
2, y2) (step 420). afterwards,
A rectangle having the area indicating points P1 and P2 as diagonal lines is formed on the image data, and coordinate values (x1, y1), (x1, y2), (x2, y1), (x1, y2) of four points of a corner portion of the rectangle are formed. (X2,
y2) is calculated (step 430).

Subsequently, an area circumscribing the image component included in the formed rectangle is extracted (step 44).
0). Specifically, x of the image part contained in the rectangle
By calculating a minimum value Xmin and a maximum value Xmax in the coordinate direction and a minimum value Ymin and a maximum value Ymax in the y coordinate direction, and extracting a rectangle having these values as four corners, the circumscribed area of the image component can be calculated. Extract. Then, step 43
The value of (y2-y1) / (x2-x1) is calculated from the coordinate values calculated at 0 and substituted into the variable T (step 450).

Thereafter, it is determined whether or not the variable T is smaller than a negative value of a predetermined value th (for example, value 1) (step 4).
60). Here, if it is determined to be smaller, the variable MO
The value 3 is substituted for D (step 462). On the other hand, the variable T
Is smaller than the negative value of the predetermined value th, it is determined whether the variable T is smaller than 0 (step 470). Here, when it is determined that the value is smaller, the value 1 is substituted for the variable MOD (step 472).
On the other hand, when it is determined that the variable T is not smaller than the value 0,
Further, it is determined whether or not the variable T is smaller than a predetermined value th (step 480). Here, when it is determined that the value is smaller, the value 2 is substituted for the variable MOD (step 482).
On the other hand, when it is determined that the variable T is not smaller than the predetermined value th, the value 4 is substituted for the variable MOD (step 48).
4).

That is, according to the processing of steps 460 to 484, when T <-th, the variable MOD has a value of 3, when -th≤T <0, when the variable MOD has a value of 1, and when 0≤T <th. The variable MOD has a value of 2 and the variable MOD has a value of 4 when th ≦ T. Thereafter, the processing shifts to “return”, and the image component extraction processing ends once.

Here, when T <-th, -th ≦ T <
The meaning of 0, 0 ≦ T <th, and th ≦ T will be described next. As shown in FIG. 7A, a rectangle formed on the basis of the region indicating points P1 and P2 is a horizontally long rectangle, and a diagonal line of the rectangle connecting both the indicating points P1 and P2 is lower right (= left upper). In this case, (y2-y1) / (x2-
The value of the variable T determined by x1) is equal to or more than -th and less than 0, and the value of the variable MOD is 1.

As shown in FIG. 7 (b), a rectangle formed based on the region pointing points P1 and P2 is a horizontally long rectangle, and a diagonal line of the rectangle connecting both the pointing points P1 and P2 is lowered to the left (= In the case of (upward right), the value of the variable T is equal to or greater than 0 and less than th, and the variable MOD is equal to 2. FIG.
As shown in (c), the rectangle formed on the basis of the region pointing points P1 and P2 is a vertically long rectangle, and the two pointing points P1 and P2
If the diagonal of the rectangle connecting
The value of the variable T is less than -th, and the variable MOD has a value of 3
Becomes As shown in FIG. 7D, the region designation points P1, P
2 is a vertically long rectangle, and the diagonal line of the rectangle connecting the two designated points P1 and P2 is lower left (= right upper), the value of the variable T is equal to or greater than the value th, and the variable MOD is The value is 4.

From the above, the region indicating point P
Four modes can be selected according to the instruction method of 1, P2. Therefore, the operator determines whether the image component to be extracted is the upright direction, the inverted direction, the rightward direction, or the leftward direction (see FIG. 8), and associates each direction with each of the four modes. Area indication points P1, P
By inputting 2, the direction of the image component can be specified at the same time as the specification of the area. Even if the boundary value of each mode is appropriately changed so as to belong to the other adjacent mode, there is no problem in the processing.

That is, when the image part to be extracted is in the erect direction, the diagonal line of the rectangle connecting the area designating points P1 and P2 becomes a right-downward rectangle as shown in FIG. By inputting the two designated points P1 and P2, the value 1 meaning erect can be stored in the variable MOD. When the image component is in the inverted direction, FIG.
A horizontally long rectangle as shown in FIG.
By inputting the two designated points P1 and P2 so that the diagonal line of the rectangle connecting P2 becomes lower left, the variable MO
D can be stored with a value of 2 which means inverted. When the image component is in the rightward direction, the input of both pointing points P1 and P2 is performed so that the diagonal of the rectangle connecting the area pointing points P1 and P2 is a vertically long rectangle as shown in FIG. By doing so, the value 3 meaning rightward can be stored in the variable MOD. When the image component is directed leftward, the two pointing points P1 and P2 are vertically elongated rectangles as shown in FIG. 7D so that the diagonal of the rectangle connecting the area pointing points P1 and P2 becomes lower left. If an input is made, the value 4 meaning leftward can be stored in the variable MOD.

When the image parts are extremely vertically or horizontally long, the MODs 1 and 3 and the MODs 2 and 4 may be confused with each other.
It is necessary to designate the region designation points P1 and P2 such that the bottom side (= upper side) of the rectangular area when viewed upright is longer than the other sides. At this time, if the character spacing is particularly narrow and the designated points of P1 and P2 are restricted by themselves in relation to other image parts,
The operator may input and specify the MOD value in that case from the keyboard.

When the image component extraction processing is completed,
Recognition processing of what characters, symbols or graphics the extracted image parts are is performed by a general-purpose pattern recognition processing (step 500).

In the recognition process, as shown in FIG.
It is determined whether the variable MOD is value 1, value 2, value 3, or any other value (steps 510, 520, 530).
If it is determined in step 510 that the value is 1, it means that the extracted image component is in the erect direction, and the process proceeds to step 570 as it is.

On the other hand, if the variable MOD is determined to be the value 2 in step 520, it means that the extracted image part is in the inverted direction, so that the entire circumscribed area of the image part extracted in step 440 is Rotate 180 degrees (step 540). If it is determined in step 530 that the variable MOD has the value 3, it means that the extracted image component is in the rightward direction. Therefore, the entire circumscribed region of the image component extracted in step 440 is rotated counterclockwise. (Step 550). If it is determined in step 530 that the variable MOD is not the value 3, that is, the value is 4, it means that the extracted image component is in the leftward direction. Rotate the entire area 90 degrees clockwise (step 5
60). After execution of step 540, 550 or 560, the process proceeds to step 570.

FIG. 8 schematically shows the contents of the processing of steps 510 to 560. As shown in the figure,
According to the processing of steps 510 to 560, when the image component is in the upright direction, the entire circumscribed region of the image component is kept without rotation, and when the image component is in the inverted direction, the entire circumscribed region is rotated by 180 degrees. When the image component is in the rightward direction, the entire circumscribed region is rotated counterclockwise by 90 degrees, and when the image component is in the leftward direction, the entire circumscribed region is rotated clockwise by 90 degrees.

Incidentally, the rotation processing of the image data is as follows.
Affine transfo, a geometric transformation process
rmatinon), and a detailed description of the affine transformation will be omitted here as it refers to, for example, “Basic Techniques for Image Processing” published by Technical Review Company.

In step 570, processing is performed to create feature data in which various features such as the linear density of the image component are expressed as mathematical expressions, from the entire circumscribed region where the image component is erected. Then, the created feature data is
A pattern recognition process for collating with a standard pattern stored in the dictionary 16a corresponding to feature data such as a figure is performed (step 580).

After the execution of the collation processing in step 580, the processing shifts to "return", and the recognition processing shown in FIG. Upon completion of the recognition process, the process proceeds to step 58.
The comparison result of 0 is output to the magnetic disk device 16 or the CRT display 14b of the interactive terminal 14 (step 600). The matching result that is output is a recognition code that identifies the standard pattern when there is a standard pattern that matches the image part. On the other hand, when there is no matching pattern, an error indicating that recognition was not possible Code. Thereafter, the pattern recognition work ends at “END” and ends.

According to the pattern recognition apparatus of the present embodiment configured as described above, the processing of steps 410 to 430 corresponding to the area designating step S1 and the steps 440 to 484 corresponding to the image component extracting / direction calculating step S2. By executing the processing of (1) and (2), the two area indication points P1 and P
2 from the mouse 14a, a variable MO indicating the direction of the image part is obtained at the same time as extracting the area circumscribing the image part.
D can also be determined. Therefore, when the direction of the image component is required in another process after the extraction of the image component, it is not necessary to input the direction of the image component again, and the operability is excellent.

In the case of the present embodiment, the direction of the image component is set to the variable M when comparing the image component with the standard pattern.
Since it is configured to know from the OD and convert the image component into an erect state and then perform the matching, the standard pattern stored in the dictionary 16a is only one erect one-way pattern for one image component. Thus, the capacity of the dictionary 16a, that is, the used capacity of the magnetic disk device 16 can be reduced.

In the above embodiment, the character "A" has been described as an example of an image part. However, the image part may of course be a symbol such as "￥" or "＄". The figure may be a figure such as the crown symbol SI shown in FIG. 6 or an aggregate thereof. Further, in the above-described embodiment, the extracted image parts are configured to be converted into the erect state and then the comparison is performed. However, instead of this, the direction of the extracted image parts is kept as it is, The dictionary of the standard pattern to be collated according to the variable MOD representing the direction may be selected from four types of dictionaries prepared in advance, erect, inverted, rightward, and leftward, and the matching process may be performed. Good.

Although one embodiment of the present invention has been described above, the present invention is not limited to such an embodiment, and the present invention is not limited to this embodiment. It goes without saying that the present invention can be carried out in various modes without departing from the gist of the invention.

[0038]

As described above in detail, according to the image data extracting method of the present invention, the direction of the image component is simultaneously detected with the extraction of the image component only by designating the rectangular area including the image component at two points. Can also be requested. Therefore, when the direction of the image component is required in another process after the extraction of the image component, it is not necessary to input the direction of the image component again,
Excellent operability.

[Brief description of the drawings]

FIG. 1 is a flowchart illustrating a basic configuration of the present invention.

FIG. 2 is a block diagram showing a schematic configuration of a pattern recognition device to which an image data extraction method as one embodiment of the present invention is applied.

FIG. 3 is a flowchart showing a pattern recognition operation performed by a processing unit 10 of the pattern recognition device.

FIG. 4 is a flowchart showing an image part extraction process in the pattern recognition work.

FIG. 5 is a flowchart showing a recognition process in the pattern recognition work.

FIG. 6 is an explanatory diagram showing an area designated point input by an operator in the pattern recognition work.

FIG. 7 is an explanatory diagram showing four modes of how to designate an area designated point.

FIG. 8 is an explanatory diagram illustrating a method of converting image components in different directions into image components in an erect direction.

[Explanation of symbols]

 S1 area designation step S2 direction calculation step 1 CPU 10 processing unit 12 image scanner 14 interactive terminal 14a mouse 16 magnetic disk device 16a dictionary P1, P2 area designation point

Claims (2)

(57) [Claims] 1. An image data extraction method for extracting an image component from raster data of an image including the image component, the method comprising: extracting a rectangular region including the image component in the raster data into two ends of one diagonal of the rectangle. An area designating step of designating two points, and a line connecting the two points serving as a basis for designating the rectangular area while extracting the image component from the rectangular area designated in the area designating step An image component extraction / direction calculation step of calculating a direction of the image component based on a minute inclination . 2. The method according to claim 1, wherein the direction of the image component is predetermined.
With a configuration that determines which of the four directions corresponds to
The image data extraction method according to claim 1.