JP3368511B2 - Apparatus and method for extracting connected area - Google Patents

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a connected area extracting apparatus and method for extracting connected areas connected in a binary image and creating label image information based on the connected areas.

In recent years, a labeling method for extracting connected regions has become important as a means for analyzing an image and extracting characters, figures, etc. from the image. For example, in a handwritten character recognition device (optical character reader: OCR), labeling is used in the preprocessing unit as a means for cutting out individual characters from a plurality of characters. Furthermore, it has been widely implemented in image processing application systems for inspection and monitoring.

[0003]

2. Description of the Related Art FIG. 23 is a diagram for explaining a conventional method. Throughout the drawings, the same reference numerals indicate the same or similar components.

In the conventional method, the labeling processing unit converts the connected binary image data from the connected component (1 in the binary image).
The components of the image that are connected to each other (three connected components exist in the figure) are calculated, and a label (1, 1,
The label images with the labels (2, 3) are stored and stored in the label image storage unit. By utilizing them in the subsequent processing, for example, image analysis and document image processing are performed. Regarding this labeling processing, for example, “Raster scanning type labeling processing method” (Japanese Patent Application No. 03-206574),
"Labeling processing method" (Japanese Patent Application No. 05-237552)
Is known.

[0005]

In the conventional method, since label image data in which connected components are labeled from the input binary image data is stored, the storage capacity required for the label image is equal to the number of connected components of the binary image. Increase proportionally. For example, if 255 labels are attached, each pixel requires 8 bits, and since the input binary image has 1 bit / pixel, the data amount becomes 8 times the number of pixels of the entire image. That is, assuming that the number of pixels in one image is n, a storage capacity of 8n bits is required.

In general image processing, the number of labels is usually 2
Considerably more than 55, for example 64K (K = 102
4) or 4G (G = 1024 ³ ). At this time, 1
16 bits or 32 bits are required for each pixel, and accordingly, the data amount is 16 times and 32 times the number of pixels of one image as a whole, respectively.

As described above, according to the conventional method, the input 2
There is a problem that a huge storage capacity is required when performing the labeling process from the value image. In particular, when the labeling process is implemented on a general-purpose processor, the storage capacity is insufficient,
There was a problem that the processing speed was significantly reduced.

It is an object of the present invention to provide a connected area extracting apparatus and method capable of economically outputting label image information to an image processing apparatus for image analysis, document image processing, etc. with a small storage capacity. And

[0009]

FIG. 1 shows a block diagram of the principle of the present invention. The binary image storage means stores the input binary image data, and the circumscribed rectangle calculation means calculates the circumscribed rectangle of the connected component from the binary image data, labels each circumscribed rectangle, and stores the circumscribed rectangle. The determining means determines whether or not each circumscribing rectangle calculated by the circumscribing rectangle calculating means is overlapped with another circumscribing rectangle, and the labeling means performs labeling processing on the circumscribing rectangle determined to be overlapped by the overlap determining means. , The means for creating the label image and the subdivision means subdivide the circumscribed rectangle of the label image created by the labeling means into the circumscribed rectangle of the child based on the label contained inside, and further circumscribe the child. Repeatedly subdivide the rectangle until the number of labels contained inside becomes single, and find the coordinates of the circumscribed rectangle of each child when the number of labels becomes single. Means, output means 2
Output data consisting of the binary image stored in the value image storage means, the coordinates of the circumscribing rectangle determined to be non-overlapping by the overlap determining means, and the circumscribing rectangle of the parent and each child determined by the subdivision means. Is a means to do.

[0010]

In the image processing apparatus which performs the image processing based on the label image information, the connected area of the pixels is extracted from the input binary image data of the screen and the information is output as the label image information of one screen. In the connected area extraction device,
The binary image storage means stores the inputted binary image data, and the circumscribed rectangle calculation means calculates the circumscribed rectangle of the connected component from the binary image data, labels each circumscribed rectangle and stores it. The overlap determination means determines whether or not each circumscribed rectangle calculated by the circumscribed rectangle calculation means overlaps with other circumscribed rectangles, and the labeling means performs labeling processing on the circumscribed rectangles determined to have overlap by the overlap determination means. ， Create the label image. The subdivision means subdivides the circumscribed rectangle of the label image created by the labeling means into the circumscribed rectangle of the child based on the label contained therein, and further divides the circumscribed rectangle of the child into the number of labels contained inside. Repeat the subdividing operation until the number of labels becomes one, and find the coordinates of the circumscribed rectangle of each child when the number of labels becomes unity. The output means is based on the binary image stored in the binary image storage means, the coordinates of the circumscribing rectangle determined to have no overlap by the overlap determining means, and the circumscribed rectangle of the parent and each child determined by the subdivision means. Output data consisting of Therefore, the image processing device can perform the image processing by reproducing the label image information of one screen based on the data output from the output means.

[0011]

FIG. 2 is a schematic diagram for explaining the principle of the present invention. (1) From the input binary image, the connected component, which is the component of the image connected to one in the binary image, is extracted. (2) When the extracted connected component further contains other connected components, each connected component (2-2, 2-3) containing the extracted connected component (2-1) inside Subdivided into regions. (3) Determine which connected component each subdivided region is based on the label. (4) Sum of regions obtained by subdividing the extracted connected component,
It is expressed by an arithmetic expression such as a difference and stored.

As described above, unlike the conventional method in which the label image of the entire screen or the label image of the connected component having the overlapping portion with other connected components is stored, the present invention is the sum of the regions where the connected component is subdivided. , The number of label images to be stored is reduced or eliminated by representing the difference.

FIG. 3 is a flow chart for explaining the outline of the first embodiment of the present invention. (1) Labeling processing [“Labeling processing method” (Japanese Patent Application No. 05-237552)] is applied to the input binary image data, and a label for distinguishing them is attached for each connected pattern (connected component). The circumscribed rectangle of the connected pattern is calculated. (The meaning of the circumscribed rectangle is the coordinates on the screen and /
Or it may include a label as its attribute. (2) Determine whether the circumscribed rectangles of each label overlap each other. (3) For the non-overlapping labels based on the judgment result, the correspondence between the non-overlapping labels (number) and its circumscribed rectangle is stored. (4) For overlapping labels, the re-circumscribing rectangle in the overlapping label-of-interest label calculates the circumscribing rectangle of each concatenated pattern again for the image in the label of interest. (5) Since the circumscribed rectangle of the label of interest matches the image size for re-circumscribed rectangle calculation, the circumscribed rectangles of labels other than the overlapping label of interest (hereinafter, this circumscribed rectangle is referred to as a noise circumscribed rectangle) are calculated. . (6) It is confirmed whether or not there is one connecting pattern included in the noise circumscribing rectangle. Specifically, the circumscribed rectangle of the connected pattern is calculated again for the image in each noise circumscribed rectangle, and whether or not there is a single label is confirmed from the number of circumscribed rectangles. (7) If all noise circumscribed rectangles are single labels, those noise circumscribed rectangles are overlapped and stored as noise rectangles in the label of interest. In this way, when the label image is restored, the area obtained by removing the noise circumscribing rectangle from the circumscribed rectangle of the overlapping target label becomes the overlapping target label area. (8) If at least one noise circumscribed rectangle is not a single label (that is, the same label as the target label), the label image in the overlapping target label is stored. This label image is calculated by performing labeling processing on the rectangular image of the overlapping label of interest. When calculating the circumscribed rectangle in (4) above,
The label images may be stored at the same time. As described above, in the present embodiment, not only the circumscribing rectangle but also the label image is represented and the connection pattern is output. Based on the circumscribed rectangle and the label image output in this way, the image processing device performs image analysis, document image processing, and the like.

FIG. 4 is a block diagram for explaining the first embodiment of the present invention. Details will be described below with reference to FIG. The labeling processing unit 2a executes the labeling processing based on, for example, the "labeling processing method" described above. Examples of the processing results are shown in FIGS. 5, 6 and 7. FIG. 5 is a diagram for explaining a label image, and each connected component in the binary image is numbered (labeled) in order from the upper left so that the connected components can be distinguished. FIG. 6 shows a label number storage area 22a for storing the total number of labels, and stores the maximum value of the assigned label numbers, that is, the maximum label value. FIG. 7 shows the circumscribed rectangle storage area 23a, in which the coordinate data of the upper left and lower right of the rectangle circumscribing each label is stored.

The circumscribed rectangle overlap determination unit 3a is a flowchart of FIG. 15 (shows an example of a circumscribed rectangle overlap determination method).
According to the above, it is determined whether the circumscribed rectangles overlap with each other based on whether or not the vertices of the circumscribed rectangle of the other party are included in the circumscribed rectangle of interest. FIG. 8 shows the circumscribing rectangle overlapping table 31a, which stores the presence or absence of overlapping of the circumscribing rectangles of the respective labels and the label of the overlapping partner in a table format.

The non-overlapping part circumscribing rectangle storage unit 4a stores only non-overlapping circumscribing rectangles in the non-overlapping part circumscribing rectangle storage area 41a (see FIG. 9) using the created circumscribing rectangle overlapping table 31a.

The overlapping part re-circumscribing rectangle calculating unit 5a first calculates the circumscribing rectangle of each connected component again in the overlapping part region, and stores it in the overlapping part re-circumscribing rectangle storage region 51a (see FIG. 10 for details).
(See Reference). For example, for label 3 in FIG. 11, upper left coordinates (X131, Y131) and lower right coordinates (X231,
The circumscribed rectangle of Y231) and the circumscribed rectangle of upper left coordinates (X132, Y132) and lower right coordinates (X232, Y232) are stored.

The noise circumscribing rectangle calculating unit 6a outside the overlapping portion attention label matches the circumscribing rectangle of the attention label stored in the circumscribing rectangle storage area 23a with the circumscribing rectangle stored in the overlapping portion recircumscribing rectangle storage area 51a. Find a circumscribed rectangle other than the label of interest by examining. If they do not match, the circumscribing rectangle other than the target label is set, and the column of the target label in the overlapping part recircumscribing rectangle storage area in FIG. 10 is set to No. For label 3, upper left coordinates (X132, Y132) and lower right coordinates (X232, Y23)
The circumscribed rectangle of 2) corresponds to this.

Single label confirmation processing unit 7a in a rectangle circumscribing noise
Calculates the circumscribed rectangle of the connected component again in the circumscribed rectangle other than the label of interest, and if the connected component is one, the single label column of the overlapping part recircumscribed rectangle storage area in FIG. To do. In Label 3 in Figure 11, there are two labels 4 and 3 in the circumscribing rectangle of interest [upper left coordinates (X132, Y132) and lower right coordinates (X232, Y232)], so the single label column is No. Become. On the other hand, for label 5 in FIG.
Since there are only 6 identical labels in the circumscribed rectangles other than the label of interest [the upper left coordinates (X152, Y152) and the lower right coordinates (X252, Y252), there is one connected component and the single label column is Yes.
Becomes

The circumscribed rectangle storage unit 8a of the overlapping attention label is
The circumscribing rectangle of the overlapping attention label is stored in the overlapping portion circumscribing rectangle storage area 89a (see FIG. 13), and the circumscribing rectangles other than the attention label of which the single label column is Yes in the table of FIG. 10 are stored.

A label image storage unit 8a for overlapping attention labels
Is stored in the label image storage area 89a (see FIG. 14) for the target label of which the single label column is No in the table of FIG. 10 (see FIG. 14), and in the overlapping portion circumscribed rectangle storage area of FIG. The head address of the storage location where the label image is stored is stored. The label image of the overlapping portion is cut out from the label image storage work area 21a by the circumscribed rectangle of the overlapping portion, and the labels other than the label of interest are stored as 0 (see FIG. 14, label image (1) to be stored). . Furthermore, in order to compress the data amount of the label image. The label of interest is 1
Alternatively, bits other than the above may be set to 0 and stored in 1 byte or 1 word by packing bits.

After performing the above processing, finally, as shown in FIG.
Only the data shown at 9, 13, and 14 are stored (it is assumed that binary images are stored naturally). When the label image is restored using these stored data, the non-overlap portion is cut out from the binary image using the coordinate values of the circumscribed rectangle in FIG. 9, and 1 is replaced with the label of interest. Also,
When the head address of the label image and the column of the single-label circumscribing rectangle are “none”, the overlapping portion is restored in the same manner as the non-overlapping portion. When there is a single-label circumscribed rectangle, the pixel value in the single-label rectangle is set to 0, in addition to the restoration similar to the non-overlapping part. When the head address of the label image is present, the coordinate values of the circumscribing rectangle in FIG. 9 and the label image (1) in FIG. 14 are used for restoration. If compressed with label image (2),
Replace 1 with the label of interest and restore the label image.

Here, the determination of whether or not the circumscribed rectangles overlap by the circumscribed rectangle overlap determination unit 4a will be described with reference to the flowchart of FIG. The determination method is performed based on whether or not the vertices of the circumscribed rectangle j of the opponent are included in the circumscribed rectangle i of interest. First, i is fixed to 1, j is changed to j = 1 to m (m is the total number of circumscribing rectangles), and whether or not the vertices of the circumscribing rectangle j are included in the circumscribing rectangle i is determined for each j. judge. Next, i is increased by 1 (i = 2), j is changed from j = 1 to m, and the same judgment is performed. The above processing is performed for i = 1 to m. When a predetermined judgment result is generated in this process, the judgment is performed by exchanging (reversing) i and j, that is, based on whether or not the vertex of the circumscribed rectangle i is included in the circumscribed rectangle j. Be seen. (S01) i is set to an initial value of 1, and first, a circumscribing rectangle j (j =
1 to m) is started by determining whether or not the vertices 1 to m are included in the circumscribed rectangle 1. (S02) The value obtained by adding 1 to i is set to j (beginning with j = 2) to determine whether or not the circumscribed rectangle i and the next circumscribed rectangle j overlap. (S03) When the overlap is determined, the arithmetic expression of inflag is executed.

Here, based on (a) of FIG. 16, the arithmetic expression in
The definition of flag is explained. inflag = in (X1i, Y1i, X2i, Y2i, a, b) is the xy coordinates of the upper left and lower right vertices, respectively (X
1i, Y1i), (X2i, Y2i) The positional relationship (inside, apex, boundary, outside) of the point (a, b) with respect to the rectangle is 16
It is indicated by two decimal digits (however, OX indicates that the following numerical values are displayed in hexadecimal, and F indicates hexadecimal 15). That is, X1i≤a, a≤X2i, and Y1i≤b, and
b ≦ Y2i [that is, the point (a, b) is within the rectangle or on the boundary thereof], and X1i <a and a <X2i and Y1i <b, and
When b <Y2i, inflag = OX20 (internal).

If the above condition is not satisfied, (X1i = a,
And Y1i = b) or (X2i = a and Y1i = b) or (X1i = a and Y2i = b) or (X2i = a and Y2i = b), inflag = 0X02 (vertex) ).

When neither of the above conditions is satisfied, infl
Set ag = 0X08 (boundary). When the above conditions are not met, inflag = 0X00 (external).

However, the inside is a point inside the rectangle and does not include the outline of the rectangle. The vertices are the four vertices of the rectangle, the boundaries are on the contour line other than the vertices, and the outside is points outside the rectangle.
Based on the above definitions, four arithmetic expressions (inflag) are sequentially executed. The first term of each arithmetic expression is the four circumscribed rectangles j of the partner circumscribed rectangle j with respect to the circumscribed rectangle i whose coordinates of the upper left and lower right vertices of the circumscribed rectangle i of interest are (X1i, Y1i) and (X2i, Y2i). Vertices (X1j, Y1j), (X2j, Y1j),
The positional relationship between (X1j, Y2j) and (X2j, Y2j) is determined. (S04) inflag of the final result of the operation and a 2-digit hexadecimal number OF (F is
Represents the decimal number 15). As a result of the comparison, when the former is larger than the latter (yes), it indicates that there are one or more internal points or two or more boundary points, and the processing of (S09) is performed. (For example, if there is one internal point,
0X20 + 0X00 + 0X00 + 0X00 = 0X20, when there are two boundary points in
flag = 0X08 + 0X00 + 0X00 + 0X08 = 0X10) (S05) As a result of the comparison in (S04), if the former is not larger than the latter (no), it is checked whether the value of inflag is equal to 0X08 or 0X02. If the result of the inspection is equal to any of them, it means that there is one boundary point or one vertex, and the process of (S06) is performed. (S06) In order to determine whether each vertex of the circumscribed rectangle i is included in the circumscribed rectangle j, i and j are exchanged, that is, the inflag inverse operation shown in (b) of FIG. 16 is performed. . Next, the inflag of the operation result is compared with 0X1F. As a result of the comparison, if the former is larger than the latter (yes), it indicates that there is one or more internal points, and the process of (S09) is performed. If the former is not larger than the latter (no), the process of (S10) is performed. (S07) As a result of the comparison in (S05) above, inflag is 0X02 even if it is 0X08.
If not equal to (no), inflag is compared with 0X04. As a result of the comparison, if the former is not larger than the latter (no), execute (S10), and if the former is larger than the latter (yes),
Execute (S08). (S08) To determine whether each vertex of the circumscribing rectangle i is included in the circumscribing rectangle j, i and j are reversed to perform the inflag inverse calculation. Next, the inflag of the operation result is compared with 0XOF. If the former is larger than the latter (yes), it indicates that there are two or more boundary points, and the process of (S09) is performed. If the former is not larger than the latter (no), the process of (S10) is performed. (S09) The circumscribing rectangle overlapping table 41 is registered with the circumscribing rectangles i and j with overlapping. (S10) In order to check whether or not the next circumscribed rectangle j overlaps the circumscribed rectangle i, 1 is added to j. (S11) The j added in (S10) is compared with the total number m of circumscribing rectangles. If j is less than or equal to m (no), then (S
Returning to 03) (the next circumscribing rectangle j is set as the circumscribing rectangle of the other party), the overlap determination is performed. When j is larger than m (yes),
It indicates that the determination process has been completed up to the last circumscribed rectangle for the circumscribed rectangle j of the other party, i is incremented by 1, and the determination process is continued while paying attention to the next circumscribed rectangle i. (S12) The i added in (S11) is compared with (m-1). When i is (m-1) or less (no), the above
Returning to (S02), overlapping is determined (the next circumscribing rectangle i is the circumscribing rectangle of interest). i is larger than (m-1) (yes
In the case of s), it indicates that the determination process has been completed to the end for the circumscribing rectangle i of interest (and by the comparison in (S11) above, the determination process has been completed to the end for the circumscribed rectangle j of the other party. ,) Overlap determination processing ends.

Next, a second embodiment of the present invention will be described. FIG. 17 is a block diagram for explaining the second embodiment of the present invention. In the first embodiment (see FIG. 4), the label images are once calculated by the labeling process, and then the presence or absence of the overlap between the label images is inspected. ) Data compression of the label image was performed. Therefore, the work area 21a for the label image is required. The second embodiment aims to reduce the work area 21a.

In FIG. 17, the label image storage work area 21a is eliminated as compared with the first embodiment (see FIG. 4). That is, the labeling processing unit 2b calculates only the number of labels and the circumscribed rectangle of each label, and the label number storage area
22b and the circumscribed rectangle storage area 23b. The following units perform the same processing as those in the first embodiment.

After that, the label image storage unit 9b of the overlapping attention label uses the binary image 1b and the data of the circumscribing rectangle storage area 23b and the circumscribing rectangle overlapping table 31b, and
A binary image in the circumscribed rectangle (denoted by α) of the overlapping portion is calculated. Next, the labeling process is performed again on the image to obtain the label image and the circumscribed rectangle. Here, since other labels are also included, the circumscribed rectangle of each label is calculated, and α
Only the label of the circumscribing rectangle that matches is retained and stored in the format of label image (2) in Fig. 14. Here, the work area of the label image to be used is the label image storage area 89b (for details, see FIG.
(See) can be substituted. As a result, in the second embodiment, the same data as in the first embodiment can be obtained with a small work area.

Next, a third embodiment of the present invention will be described. FIG. 18 is a block diagram for explaining the third embodiment of the present invention. In each of the examples of Embodiments 1 and 2 (see the block diagrams of FIGS. 4 and 17), the circumscribed rectangle of the attention label is calculated as the circumscribed rectangle only when the circumscribed rectangle other than the attention label is single at the overlapping portion. Difference), and if there are multiple circumscribing rectangles other than the label of interest at the overlapping part, the label image is stored. In the third embodiment, in order to reduce the amount of data required to store the label image, the subdivision operation is repeated until the number of labels included in the circumscribed rectangle (called noise circumscribed rectangle) other than the target label becomes single. That is, after performing the same processing as in the second embodiment, the overlapping portion re-circumscribing rectangle calculation unit
5c, the noise circumscribing rectangle calculation unit 6c outside the label of interest in the overlapping portion, and the single label confirmation processing unit 7c in the noise circumscribing rectangle repeat the subdivision processing until the number of labels included in the noise circumscribing rectangle becomes single.

19 to 21 show a processing example thereof. First, the figure
As shown in FIG. 19, the process of subdividing the noise circumscribed rectangle is repeated until the label in each circumscribed rectangle (noise circumscribed rectangle) in the circumscribed rectangle of the target label becomes single. That is, when the circumscribed rectangle of the label of interest is the parent, the circumscribed rectangle of the label of interest can be expressed hierarchically, such as child 1, child 2, ... In the overlapping portion re-circumscribing rectangle storage area 51c (see FIG. 20 for details), which is stored by the noise circumscribing rectangle calculating unit 6c outside the overlapping portion attention label, the rectangle attribute 1 of the child 1 is the first label of the hierarchy of the child 1, The rectangle attribute 1-1 of child 2 means the first label belonging to the first rectangle of child 1. In the overlapping attention label circumscribing rectangle storage area 89c (see FIG. 21 for details) stored by the circumscribing rectangle storage unit 8c of the overlapping attention label,
It is determined whether each circumscribed rectangle subdivided into a single label is the label of interest, and if it is the label of interest, the operation attribute is "+",
If not, set it to "-". Naturally, all the operation attributes of the hierarchy of child 1 are "-". To determine whether the label is the target label, the circumscribed rectangle of the target label is labeled to calculate a label image, and it is determined whether the label value in the circumscribed rectangle of the child is equal to the label value of the parent. In the method of restoring the overlapping portion, the label image is obtained by calculating the area of each label from FIG. 21 and replacing each pixel of the pixel value 1 in that area in the binary image with the label number. For example,
As shown in FIG. 21, the area of label number 9 can be restored by performing addition and subtraction using the operation attribute in the following order from parent to child 1, child 2 ...

When performing the subdivision processing of the overlapping portion, if the area is large or the number of labels is large, there is a problem that the work area for temporarily storing the label image of the overlapping portion becomes large. Therefore, the problem can be solved by dividing the target area and calculating the label image. For example, when calculating the calculation attribute of FIG. 21, a label image is calculated using the circumscribed rectangle of the parent as the target area to determine whether or not the label is the target label.
The label image is calculated by dividing as in (a). At this time,
If the target area is vertically long, it is divided horizontally, and if it is horizontally long, it is vertically divided. Of course, the circumscribed rectangle of the label that determines the operation attribute should be included. When the label image is calculated in the divided area, a label having a circumscribed rectangle having the same size as the divided area becomes the target label of the target area. if,
If there is no label having a circumscribed rectangle of the same size as the divided area, the divided area is set again. Figure 22
The label image can be calculated in a small work area as shown in (b).

As described above, the division area setting method divides horizontally when the target area is vertically long and vertically when the target area is horizontally long, and includes the circumscribed rectangle of the label for determining the operation attribute. ， Set to the area larger than the circumscribed area of the related label. Then, the label image can be calculated by enlarging the divided area until a label having a circumscribed rectangle having the same size as the divided area appears.

[0035]

As described above, according to the method for extracting a connected area of the present invention, an image processing apparatus for image processing based on label image information includes coordinates and labels of a circumscribed rectangle having no overlap. In addition to outputting the circumscribed rectangle information, (1) For circumscribed rectangles that have overlap, subdivide them into child circumscribed rectangles, and if there are multiple labels for the child circumscribed rectangles contained in the circumscribed rectangle, the parent The label image of the circumscribing rectangle is output, and when the label is single, the coordinates of the circumscribing rectangle of the parent and child are output to the image processing device. Or
(2) For a bounding rectangle that has an overlap, the operation of subdividing the bounding rectangle as a parent into the bounding rectangle of the child is repeated until the label of the bounding rectangle of the child contained within is repeated until each child at that time. The coordinates of the circumscribed rectangle and the attribute indicating whether or not the rectangle belongs to the circumscribed rectangle of the parent are output to the image processing apparatus. The image processing apparatus can easily reproduce the label image of the entire screen based on the information thus input. Therefore, the present invention can reduce or eliminate the amount of label images in comparison with the conventional method that requires a huge storage capacity to output the label image of the entire screen to the image processing unit, and the amount is extremely small. There is an effect that it is possible to realize an economical connected area extracting apparatus and method by only requiring the storage capacity of

[Brief description of drawings]

FIG. 1 is a block diagram of the principle of the present invention.

FIG. 2 is a conceptual diagram illustrating the principle of the present invention.

FIG. 3 is a flowchart for explaining the outline of the first embodiment of the present invention.

FIG. 4 is a block diagram illustrating a first embodiment of the present invention.

FIG. 5 is a diagram illustrating a label image.

FIG. 6 is a diagram illustrating a label number storage area.

FIG. 7 is a diagram illustrating a circumscribed rectangle storage area.

FIG. 8 is a diagram illustrating a circumscribing rectangle overlapping table.

FIG. 9 is a diagram illustrating a non-overlapping portion circumscribed rectangle storage area.

FIG. 10 is a diagram (part 1) for explaining a re-circumscribing rectangle storage area of an overlapping portion.

FIG. 11 is a diagram showing an example of an overlapping portion that does not have a single level.

FIG. 12 is a diagram showing an example of an overlapping portion having a single level.

FIG. 13 is a diagram (part 1) explaining an overlapping part circumscribed rectangle storage area.

FIG. 14 is a diagram illustrating a label image of an overlapping portion.

FIG. 15 is a flowchart of circumscribed rectangle overlap determination processing.

FIG. 16 is an explanatory diagram of an arithmetic expression

FIG. 17 is a block diagram illustrating a second embodiment of the present invention.

FIG. 18 is a block diagram illustrating a third embodiment of the present invention.

FIG. 19 is a diagram illustrating a subdivision process of a noise circumscribed rectangle.

FIG. 20 is a diagram (part 2) explaining the overlapping part re-circumscribing rectangle storage area.

FIG. 21 is a diagram (part 2) explaining an overlapping part circumscribed rectangle storage area.

FIG. 22 is a block diagram illustrating reduction of the label image storage work area by dividing the target area.

FIG. 23 is a diagram illustrating a conventional method.

[Explanation of symbols]

1a Binary image storage area 2a ~ 2c Labeling processing section 3a to 3c Circumscribed rectangle overlap determination section 4a to 4c Non-overlapping area circumscribed rectangle storage area 5a to 5c Overlap part Re-circumscribing rectangle calculation part 6a to 6c Overlapping part Noise circumscribed rectangle calculation part outside the label of interest 7a to 7c Single label confirmation part inside the noise circumscribed rectangle 8a to 8c Overlapping attention label circumscribed rectangle storage unit 9a, 9b Label image memory for overlapping attention labels 21a Label image storage work area 22a to 22c Label number storage area 23a to 23c Enclosing rectangle storage area 31a ~ 31c circumscribed rectangular overlapping table 41a to 41c Non-overlapping part circumscribed rectangle storage area 51a to 51c Overlapping area re-circumscribing rectangle storage area 89a, 89b Storage area for bounding box and label image 89c Overlap rectangle storage area

Front page continued (58) Fields surveyed (Int.Cl. ⁷ , DB name) G06T ⁷ /00-7/60 G06T 1/00 JISST file (JOIS)

Claims (22)

(57) [Claims] 1. An image processing apparatus that performs image processing based on label image information extracts a connected region of pixels from the input binary image data by labeling, and uses that information as label image information for one screen. In a connected area extraction device for outputting, a second circumscribed rectangle that circumscribes a connected area of a second label in a first circumscribed rectangle that circumscribes a connected area of a first label extracted from binary image data Determines that is included
First judging means and the second circumscribing rectangle based on the judgment of the first judging means.
It is determined whether the first label is included in the
The area of the first label in the second circumscribed rectangle is
The third and second determination means for determining a circumscribing rectangle, based on the determination of the second determination means, the first circumscribed rectangle in contact
Remove the circumscribed rectangular second from or circumscribing side of the first
Removing the second circumscribing rectangle from the shape and the third circumscribing
Information indicating that a square is added is added to the first label.
An output device for outputting label image information of a connected area to the image processing apparatus, the connected area extracting device. 2. An image processing apparatus that performs image processing based on label image information extracts a connected area of pixels from the input binary image data by labeling, and uses that information as label image information for one screen. In the connected area extraction device for output, the connected area of the first label extracted from the binary image data
The connecting area of the first circumscribed rectangle and the second label that circumscribe
Having a region where the circumscribing second circumscribing rectangle overlaps
First determining means for determining, and based on the determination of the first determining means, within the overlapping area
It is determined whether the first label is included, and if it is included,
Circumscribes the area of the first label within the overlapping area
Second judging means for obtaining a third circumscribing rectangle, and the first circumscribing rectangle based on the judgment of the second judging means.
From the overlapping area, or the first circumscribed rectangle
Delete the overlapping area and add the third circumscribed rectangle
The information indicating that the
Output to the image processing device as label image information
An apparatus for extracting a connected area, characterized by comprising: 3. The output means further comprises the first label.
3. The connection area extracting device according to claim 1 , wherein label image information of the connection area of the connection area is output. 4. The output means further comprises the first label.
Data in which the label of the connected region of the
Connection region according to claim 1 or 2, and outputs the number of the extraction device. 5. An image processing apparatus that performs image processing based on label image information extracts a connected region of pixels from input binary image data and outputs the information as label image information for one screen. In a connected area extracting apparatus, a binary image storage unit that stores input binary image data, and a connected component that is a component of one connected image are calculated from the binary image data, and the connected component is obtained. Labeling means for performing a labeling process to create a labeled image with labels, memorizing the label image, circumscribing rectangle calculating means for calculating and storing coordinates of the circumscribed rectangle of the connected component on the screen, and circumscribing rectangle calculation Overlap judging means for judging whether or not each circumscribing rectangle calculated by the means overlaps with other circumscribing rectangles; As a parent, a subdivision means for subdividing the child's circumscribed rectangle based on the labels contained therein and for obtaining the coordinates of the circumscribed rectangle of the child, and the number of labels of the circumscribed rectangle of the child subdivided by the subdivision means The number of labels, the binary image stored in the binary image storage means, the coordinates of the circumscribed rectangle determined to be non-overlapping by the overlap determining means, and the number of labels. Providing output means for outputting data consisting of the coordinates of the circumscribed rectangle of the parent and the child whose number of labels is judged to be single by the judgment means and the circumscribed rectangular label image of the parent whose number of labels is judged to be plural A device for extracting connected regions, characterized by: 6. An image processing apparatus that performs image processing based on label image information extracts a connected region of pixels from input binary image data and outputs the information as label image information for one screen. In a connected area extraction device, a binary image storage unit that stores input binary image data, and a circumscribed box that calculates a circumscribed rectangle of a connected component from the binary image data, labels each circumscribed rectangle, and stores the circumscribed rectangle. With respect to the circumscribing rectangles determined to have overlap by the rectangle calculating means, the overlap determining means for determining whether or not each circumscribing rectangle calculated by the circumscribing rectangle calculating means overlaps with other circumscribing rectangles, Labeling means for performing a labeling process to create and store the label image and a circumscribed rectangle of the label image created by the labeling means as a parent. A subdivision means for subdividing the child's circumscribed rectangle based on the labels contained inside to find the coordinates of the child's circumscribed rectangle, and whether the number of labels of the child's circumscribed rectangle subdivided by the subdivision means is single A label number determining means for determining whether there is a plurality, a binary image stored in the binary image storing means, coordinates of a circumscribed rectangle determined to be non-overlapping by the overlap determining means, and a label for the label number determining means. The present invention is characterized in that output means is provided for outputting data consisting of circumscribed rectangle coordinates of the parent and child whose number is determined to be single, and a circumscribed rectangle label image of the parent whose number of labels is determined to be plural. Connected area extraction device. 7. The output means outputs data and a label in which the label is replaced with 1 instead of the label image of the circumscribed rectangle of the parent whose label is determined to be plural. Alternatively, the connected area extracting device according to item 6. 8. An image processing apparatus that performs image processing based on label image information extracts a connected region of pixels from input binary image data and outputs the information as label image information for one screen. In a connected area extraction device, a binary image storage unit that stores input binary image data, and a circumscribed box that calculates a circumscribed rectangle of a connected component from the binary image data, labels each circumscribed rectangle, and stores the circumscribed rectangle. Square calculating means, overlap determining means for determining whether or not each circumscribing rectangle calculated by the circumscribing rectangle calculating means overlaps with other circumscribing rectangles, and labeling of the circumscribing rectangles determined to have overlap by the overlap determining means. Labeling means for performing the processing to create the label image and the circumscribed rectangle of the label image created by the labeling means are included in the inside as parents. Subdivide the child's circumscribed rectangle based on the label, and then further subdivide the child's circumscribed rectangle until the number of labels contained inside is repeated. Subdividing means for obtaining the coordinates of the circumscribing rectangle of the child, the binary image stored in the binary image storing means, the circumscribing rectangle determined to be non-overlapping by the overlap determining means, and the subdividing means And an output means for outputting data including the circumscribed rectangle of each parent and each child. 9. The label number determining means performs a labeling process on a circumscribed rectangle of a parent to obtain a label image of the circumscribed rectangle of the parent, and a circumscribed quadrangle label of each subdivided child. 7. The connected area extracting device according to claim 5, wherein it is determined whether the number of labels is single or plural by comparing. 10. The subdivision means performs a labeling process on a circumscribed rectangle of a parent to obtain its label image, and the label of the circumscribed rectangle of each child indicates a match / mismatch with the label of the circumscribed rectangle of the parent. 9. An attribute assigning unit for obtaining an attribute by comparing labels, the output unit outputting the coordinates of the circumscribed rectangle of each child and the attribute data obtained by the attribute assigning unit. Connected area extraction device. 11. The attribute assigning means defines a circumscribed rectangle of a parent,
The attribute is determined based on a label image obtained by performing a labeling process on the divided area so as to include the circumscribing rectangle of the child whose attribute is to be obtained.
10. The connected area extracting device according to 10. 12. An image processing apparatus that performs image processing based on label image information extracts a connected region of pixels from the input binary image data by labeling and uses the information as label image information for one screen. in the extraction method of the connection area to be output, (a) a first second communication <br/> binding region of the label into the first circumscribed rectangle which circumscribes the coupling region of the label extracted from the binary image data A step of determining that a second circumscribing rectangle that circumscribes is included, and (b) based on the determination of step (a),
Is determined whether or not the first label is included in
When the area of the first label within the second circumscribed rectangle is
A step of obtaining a third circumscribing rectangle that touches, and (c) the first circumscribing rectangle based on the judgment in the step (b) .
The second circumscribing rectangle, or the second circumscribing rectangle is deleted from the first circumscribing rectangle , and the third circumscribing rectangle .
Information indicating that a shape is added is added to the first label sequence.
And a step of outputting the label image information of the connection area to the image processing apparatus, the extraction method of the connection area. 13. An image processing apparatus that performs image processing based on label image information extracts a connected region of pixels from the input binary image data by labeling and uses the information as label image information for one screen. In the extraction method of the connected area to be output, (a) the connection of the first label extracted from the binary image data
Connecting area of the first circumscribed rectangle and the second label that circumscribe the area
It has a region that overlaps with a second circumscribed rectangle that circumscribes the region.
And (b) based on the judgment in step (a),
It is determined whether the first label is included, and when it is included.
Is a first circumscribed region of the first label within the overlapping region.
3 ) the step of obtaining the circumscribed rectangle, and (c) the first circumscribed rectangle based on the judgment of the step (b).
From the overlapping area, or from the first circumscribed rectangle
Delete the overlapping area and add a third circumscribed rectangle
Information indicating that the label of the connection area of the first label is
Outputting to the image processing device as bell image information;
A method for extracting a connected area, comprising: 14. The output of the step (c) further comprises:
Method of extracting a connected region of claim 12, wherein outputting the label image information of the connection area of the first label. 15. The output of the step (c) further comprises:
Data in which the label of the connected area of the label of 1 is replaced with 1
Claim and outputs the number of the label 12
Or a method for extracting a connected region according to 13 above. 16. An image processing apparatus that performs image processing based on label image information extracts a connected region of pixels from input binary image data and outputs the information as label image information for one screen. In the method of extracting a connected region, (a) a step of storing the input binary image data, and (b) calculating a connected component from the binary image data, and creating a label image with the connected component labeled. The process of labeling and storing the label image, (c) the process of calculating and storing the coordinates on the screen of the circumscribed rectangle of the connected component, and (d) the circumscribed tangent calculated by the process (c). A step of determining whether or not a rectangle overlaps with other circumscribing rectangles; and (e) a circumscribed rectangle of the parent determined to have overlap in step (d), and a circumscribed child of the child based on the label contained inside. Subdivided into squares, circumscribed rectangle of child And (f) determining whether the number of labels of the circumscribed rectangle of the child subdivided by the step (e) is single or plural, and (g) storing it by the step (a). Binary image and the process
The coordinates of the circumscribed rectangle determined to be non-overlapping in (d),
A step of outputting data consisting of the coordinates of the circumscribed rectangle of the parent and the child whose number of labels is determined to be single by the step (f) and the label image of the circumscribed rectangle of the parent whose number of labels is determined to be plural. A method for extracting a connected region, which is characterized by being provided. 17. An image processing apparatus that performs image processing based on label image information extracts a connected area of pixels from input binary image data and outputs the information as label image information for one screen. In the method of extracting a connected region, (a) storing the input binary image data, and (b) calculating the circumscribed rectangle of the connected component from the binary image data,
Labeling and storing each circumscribed rectangle; (c) determining whether or not each circumscribed rectangle calculated in step (b) overlaps another circumscribed rectangle; and (d) the process. Labeling processing is performed on the circumscribing rectangles that are determined to have overlap in (c), and the label image is created and stored; and (e) The circumscribed parent of the label image created in step (d). The step of subdividing the rectangle into the circumscribed rectangles of the child based on the labels contained in the rectangle to obtain the coordinates of the circumscribed rectangle of the child, and (f) the circumscribed rectangle of the child subdivided by the step (e). A step of determining whether the number of labels is single or plural; (g) a binary image stored in the step (a);
The coordinates of the circumscribed rectangle determined as non-overlapping in (c),
A step of outputting data consisting of the coordinates of the circumscribed rectangle of the parent and the child whose number of labels is determined to be single by the step (f) and the label image of the circumscribed rectangle of the parent whose number of labels is determined to be plural. A method for extracting a connected region, which is characterized by being provided. 18. The output step (g) is characterized in that, instead of the circumscribed rectangular label image of the parent whose label is determined to be plural, the label is replaced with 1 and the output data and the label are output. 18. The method for extracting a connected region according to claim 16 or 17. 19. An image processing apparatus that performs image processing based on label image information extracts a connected region of pixels from input binary image data and outputs the information as label image information for one screen. In the method of extracting a connected region, (a) storing the input binary image data, and (b) calculating the circumscribed rectangle of the connected component from the binary image data,
Labeling and storing each circumscribed rectangle; (c) determining whether or not each circumscribed rectangle calculated in step (b) overlaps another circumscribed rectangle; and (d) the process. The step of creating a label image by performing labeling processing on the circumscribing rectangle determined to have overlap in (c) and (e) the circumscribed rectangle of the parent of the label image created in step (d) are Repeatedly subdivide the child's circumscribed rectangle based on the labels contained inside, and further subdivide the child's circumscribed rectangle until the number of labels contained inside becomes single, resulting in a single label. And (f) the binary image stored in step (a), and the step of obtaining the coordinates of the circumscribed rectangle of each child
The coordinates of the circumscribed rectangle determined as non-overlapping in (c),
And a step of outputting data consisting of a parent and a circumscribed rectangle of each child obtained in the step (e). 20. In the step (f) of determining the number of labels, a labeling process is performed on the circumscribed rectangle of the parent to obtain its label image, and the label of the circumscribed rectangle of the parent and each child subdivided. 18. The method for extracting a connected region according to claim 16, wherein it is determined whether the number of labels is single or plural by comparing with the circumscribed rectangular label. 21. In the step (e) of subdividing, a labeling process is performed on the circumscribed rectangle of the parent to obtain its label image, and the circumscribed quadrangle label of each child corresponds to the circumscribed quadrangle label of the parent. There is a step (g) for finding the attribute indicating the match / mismatch by comparing the labels, and the output step in (f) above includes the coordinates of the circumscribed rectangle of each child and the attribute data found by the step (g). 20. The method for extracting a connected area according to claim 19, wherein the extraction is performed. 22. In the step (g), the circumscribed rectangle of the parent is made to include the circumscribed rectangle of the child whose attribute is to be determined, and based on the label image obtained by performing the labeling process on the divided regions, 22. Attribute determination
A method for extracting a connected region as described.