JP4322999B2 - Connected region extraction apparatus and recording medium - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
In recent years, demand for character recognition devices OCR (hereinafter simply referred to as “OCR”) is increasing as input peripheral devices for financial documents, business documents, and the like. The present invention is used for such OCR and the like, and relates to, for example, a connected region extraction apparatus and a recording medium that are effective when applied to character cutout processing in OCR.
[0002]
In the character cutout process in the OCR, in order to realize a high recognition rate of individual characters, “cutout process”, which is a stage before recognition, is important in terms of accuracy. Conventionally, characters within the specified range are specified for documents that have previously specified the character entry position such as bookkeeping (those that are written in the same color or density as the black borders and characters instead of dropout color). If written well, characters could be automatically recognized at a considerable rate, but beyond the specified range, touch the frame or ruled line indicating the specified range, or the frame or ruled line indicating the specified range. When the protrusions occur, the recognition rate is remarkably lowered.
[0003]
Therefore, there has been a demand for the development of a technique for accurately cutting only characters from characters touching ruled lines and frames. The invention of the present application is not limited to OCR, but in a broad sense as a straight line such as a printed character recognition device, cut out characters and symbols in drawing recognition, separation of ruled lines and objects, figures, and character contact portions in images. It is also possible to use the technique for extracting only the figure from the pattern in which the figures overlap.
[0004]
[Prior art]
A conventional example will be described below.
[0005]
§1: General description of the conventional example
Conventionally, there has been known a connected region extraction apparatus that extracts a connected region (connected component) in which pixels are connected to one from binary image data and creates label image information based on the connected region (for example, Japanese Patent Laid-Open No. Hei. No. 8-55219, JP-A-7-325921).
[0006]
In recent years, a labeling method for extracting a connected region where pixels are connected from binary image data has become important as means for analyzing an image and extracting characters, figures, and the like from the image. For example, in a handwritten character recognition device (OCR), a labeling technique is used for the preprocessing unit as means for cutting out individual characters from a plurality of characters. Furthermore, it has been widely realized in image processing application systems for the purpose of inspection and monitoring. The labeling technique is well known as described in, for example, the specification and drawings of Japanese Patent Application No. 3-206574 and Japanese Patent Application No. 5-237552, and the description thereof will be omitted. The outline of the invention described in JP-A-8-55219 will be described below as a conventional example.
[0007]
§2: Outline of connected area extraction device
The connected area extracting device calculates a circumscribed rectangle of a connected area (connected component) in which pixels are connected from input binary image data, labels each circumscribed rectangle with a label, and stores the circumscribed rectangle in a memory. When the circumscribing rectangle calculated by the circumscribing rectangle calculating unit determines whether or not there is an overlap with another circumscribing rectangle, and when the overlapping determining unit determines that there is an overlap, the circumscribing rectangle is again displayed. Labeling is performed to create the label image. The circumscribing rectangle of the label image is used as a parent. Based on the label contained in the label, it is subdivided into child circumscribing rectangles. The subdivision means for repeating the subdivision operation until the number of contained labels becomes single, and obtaining the coordinates of the circumscribed rectangle of each child when the number of labels becomes single, and the input Value image, configured to include no overlap and the determined circumscribed rectangle coordinates, and an output means for outputting data composed of a circumscribed rectangle of the parent and each child determined by subdivision means.
[0008]
The processing outline of the connected area extracting apparatus is as follows. The circumscribed rectangle calculating means calculates a circumscribed rectangle of the connected area from the input binary image data, and stores each circumscribed rectangle with a label. Next, the overlap determining unit determines whether each circumscribed rectangle calculated by the circumscribed rectangle calculating unit overlaps with other circumscribed rectangles. If it is determined that there is an overlap, the subdivision unit performs further labeling processing again. To perform re-division processing.
[0009]
In this case, the subdivision means uses the circumscribed rectangle of the label image that has been labeled again as a parent, and subdivides it into child circumscribed rectangles based on the labels contained therein (calculates the circumscribed rectangle again), and further The operation of subdividing the child's circumscribed rectangle is repeated until the number of labels contained therein becomes single, and the coordinates of the circumscribed rectangle of each child when the number of labels becomes single are obtained. The output means outputs data consisting of the input binary image data, the coordinates of the circumscribed rectangle determined as non-overlapping by the overlap determining means, and the circumscribed rectangle of the parent and each child obtained by the subdivision means To do.
[0010]
§3: Outline of connected area extraction processing
FIG. 8 is an explanatory diagram (part 1) of a conventional example. Hereinafter, based on the specific example of FIG. 8, the outline | summary of the connection area | region extraction process in the said connection area | region extraction apparatus is demonstrated.
[0011]
The connected component extraction unit 1 extracts a connected component (or connected region) that is a component in which pixels are connected from the input binary image data. At this time, a labeling process is performed on the extracted connected components. For example, in the illustrated example, there are a connected component of label 1 and a connected component of label 2. Next, the connected component subdivision unit 2, when another connected component is further included in the extracted connected component, converts the extracted connected component (2-1) into each connected component included therein. Subdivide into component (2-2, 2-3) regions.
[0012]
And the subdivision area | region determination part 3 determines which connected component each area | region subdivided based on the said label. In this example, the subdivision area (2-1) belongs to the connected component of label 2, (2-2) belongs to the connected component of label 1, and (2-3) belongs to the connected component of label 2. And so on.
[0013]
The subdivision area storage unit 4 expresses the extracted connected components by arithmetic expressions such as the sum and difference of subdivided areas (subdivision areas) and stores them in a storage unit (buffer memory or the like). For example, this data is stored in the storage unit in the following manner: label 2 region = + (2-1) region− (2-2) region + (2-3) region.
[0014]
As described above, unlike the method of storing the label of the entire screen, or the connected component label image that overlaps with other connected components, the connected components are expressed by the sum, difference, etc. of the subdivided areas. As a result, the number of stored label images can be reduced or eliminated.
[0015]
§4: Explanation of processing by flowchart
FIG. 9 is an explanatory diagram (part 2) of the conventional example. Detailed processing will be described below based on the flowchart of FIG. In addition, (1)-(8) of a figure shows each process.
[0016]
(1): First, the input binary image data is subjected to a labeling process (for example, refer to the specification and drawings of Japanese Patent Application No. 3-206574 and Japanese Patent Application No. 5-237552) and connected patterns ( For each connected component), a circumscribed rectangle of the connected pattern with a label for distinguishing them is calculated. In this case, the meaning of the circumscribed rectangle may include coordinates and labels on the screen as attributes.
[0017]
(2): Next, it is determined whether or not the circumscribed rectangles of each label overlap each other. That is, the overlapping of circumscribed rectangles is determined.
[0018]
(3): Based on the result of the overlap determination in (2), for the non-overlapping label (non-overlapping label), the correspondence between the non-overlapping label (number) and its circumscribed rectangle is stored in the storage unit. Remember.
[0019]
(4): For the overlapping labels, the circumscribed rectangle for each connected pattern is calculated again for the overlapping labels based on the result of the overlap determination in (2) Perform circumscribed rectangle calculation processing).
[0020]
(5): Next, since the circumscribed rectangle of the target label matches the image size for re- circumscribed rectangle calculation, the circumscribed rectangle of the label other than the overlapping target label (hereinafter, this circumscribed rectangle is referred to as “noise circumscribed rectangle”). "Rectangle" or "other circumscribed rectangle").
[0021]
(6): It is confirmed whether or not the plurality of connection patterns included in the noise circumscribed rectangle is one. Specifically, for each noise circumscribed rectangle, the circumscribed rectangle of the connected pattern is calculated again, and it is confirmed from the number of circumscribed rectangles whether or not it is a single label.
[0022]
(7): When a plurality of circumscribed rectangles are all single labels, the noise circumscribed rectangles are overlapped and stored in the storage unit as noise rectangles in the target label. Thus, when the label image is restored, an area obtained by removing the noise circumscribed rectangle from the circumscribed rectangle of the overlap target label becomes the overlap target label area.
[0023]
(8): If even one noise circumscribed rectangle is not a single label (that is, the same label as the target label), the label image in the overlap target label is stored in the storage unit. This label image is calculated by performing a labeling process on the in-rectangular image of the overlapping attention label. However, the label image may be stored at the same time when the circumscribed rectangle is calculated in (4).
[0024]
In this way, the connected pattern is output not only with the circumscribed rectangle but also with the label image. Based on the circumscribed rectangle and the label image output in this way, the image processing apparatus performs image analysis, document image processing, and the like.
[0025]
§5: Explanation of subdivision processing example
FIG. 10 is an explanatory diagram (part 3) of the conventional example. Hereinafter, an example of subdivision processing will be described with reference to FIG.
[0026]
In the above example, only when there is a single circumscribed rectangle other than the target label in the overlapping portion, the circumscribed rectangle of the target label is expressed by the circumscribed rectangle arithmetic expression, and there are a plurality of circumscribed rectangles other than the target label in the overlapping portion. In some cases, a label image was stored. However, in this example, in order to reduce the amount of data required for storing the label image, the subdivision is repeated until there is a single label included in the circumscribed rectangle (noise circumscribed rectangle) other than the target label.
[0027]
In this subdivision process, first, the process of subdividing the noise circumscribed rectangle is repeated until there is a single label in each circumscribed rectangle (noise circumscribed rectangle) present in the circumscribed rectangle of the target label. In this case, if the circumscribed rectangle of the target label is a parent (in this example, label 9), the circumscribed rectangle of the target label can be expressed hierarchically as child 1, child 2,.
[0028]
In this example, the circumscribed rectangle represented by the upper left coordinates (X19, Y19) and the lower right coordinates (X29, Y29) is the parent of the label 9, the upper left coordinates (X192, Y192), and the lower right coordinates The circumscribed rectangle represented by the coordinates (X292, Y292) is the child 1 of the label 1, and the circumscribed rectangle represented by the upper left coordinates (X193, Y193) and the lower right coordinates (X293, Y293) is the label. It is a child 2 of 1-1.
[0029]
In this way, the process of subdividing the noise circumscribed rectangle is repeated until there is a single label in each circumscribed rectangle existing in the circumscribed rectangle of the target label. Then, it is expressed by an arithmetic expression such as the sum and difference of each subdivided area (subdivided area) and stored in a storage unit (buffer memory or the like). For example, this data is stored in the storage unit in the following manner: label 9 area = + (9-1) area− (9-2) area + (9-3) area.
[0030]
[Problems to be solved by the invention]
The conventional apparatus as described above has the following problems.
[0031]
(1): In the conventional subdivision process, a label image is generated again by performing a labeling process on the circumscribed rectangle determined to be overlapped, and the label included in the circumscribed rectangle of the label image is used as a parent. Based on the above, the child's circumscribed rectangle is subdivided into sub-rectangular rectangles, and the child's circumscribed rectangle is further subdivided until the number of labels contained therein becomes single.
[0032]
For this reason, when there are small labels other than noise in the circumscribed rectangles that overlap, the number of partial labeling increases, and the time for the subdivision processing increases.
[0033]
(2): When dealing with a blurred image in the conventional connected region extraction process, since the blurred image has many fine labels, the labeling process and circumscribed rectangle calculation process are repeated many times in the subdivision process. Had to be done. For this reason, processing time was long.
[0034]
An object of the present invention is to solve such a conventional problem and to reduce the processing amount of the entire connected region extraction processing to achieve high-speed processing.
[0035]
[Means for Solving the Problems]
FIG. 1 is a diagram illustrating the principle of the present invention. In order to achieve the above object, the present invention is configured as follows.
[0036]
(1): A circumscribed rectangle calculating unit 12A that calculates a circumscribed rectangle of a connected region where pixels are connected from input binary image data, and stores each circumscribed rectangle with a label, and the circumscribed rectangle calculating unit 12A. For each circumscribed rectangle calculated, an overlap determination unit 13A that determines whether or not there is an overlap with other circumscribed rectangles, and a labeling process is performed again on the circumscribed rectangle that is determined to be overlapping by the determination of the overlap determination unit 13A. A label image is created, and the circumscribed rectangle of the label image is used as a parent, and is subdivided into child circumscribed rectangles based on the labels contained therein, and the child circumscribed rectangle is further subdivided into the number of labels contained therein. The subdivision means 15A for obtaining the coordinates of the circumscribed rectangle of each child when the number of labels becomes single, and the input binary image data; A connecting unit provided with output means 16A for outputting data consisting of the coordinates of circumscribed rectangles determined not to overlap by the determination of the overlap determining means, and the circumscribed rectangles of the parent and each child obtained by the subdivision means In the area extraction device, for the circumscribed rectangle determined to be overlapped by the determination of the overlap determining unit, the size of the circumscribed rectangle in the vertical direction or the horizontal direction is calculated, and one of the sizes is 2 pixels or less. If the determination result is a single label, coordinate data of the circumscribed rectangle is provided. If it is not a single label, it is provided with a function of sending the circumscribed rectangle coordinate data to the subdivision means 15A.
[0037]
(2): In the connected region extraction apparatus according to (1), the subdivision means 15A performs a vertical direction or a horizontal direction of the circumscribed rectangle with respect to the circumscribed rectangle calculated in the process of performing the subdivision processing of the circumscribed rectangle. A single label determining means 17A for determining that there is no other overlapping label inside the label when any of the sizes is 2 pixels or less, and If the determination result of the single label determination means 17A is a single label, the subdivision processing for the circumscribed rectangle is terminated, and if it is not a single label, the subdivision processing for the circumscribed rectangle is performed. ing.
[0038]
(3): A circumscribed rectangle calculating unit 12A that calculates a circumscribed rectangle of a connected region where pixels are connected from input binary image data, and stores each circumscribed rectangle with a label, and the circumscribed rectangle calculating unit 12A. For each circumscribed rectangle calculated, an overlap determination unit 13A that determines whether or not there is an overlap with other circumscribed rectangles, and a labeling process is performed again on the circumscribed rectangle that is determined to be overlapping by the determination of the overlap determination unit 13A. The label image is created, and the circumscribed rectangle of the label image is used as a parent. Based on the label contained in the label image, the child circumscribed rectangle is subdivided. Further, the number of labels contained in the child circumscribed rectangle is determined. The subdivision operation is repeated until the number of labels becomes single, and the subdivision means 15A for obtaining the coordinates of the circumscribed rectangle of each child when the number of labels becomes single, the input binary image data, A connecting unit provided with output means 16A for outputting data consisting of the coordinates of circumscribed rectangles determined not to overlap by the determination of the overlap determining means, and the circumscribed rectangles of the parent and each child obtained by the subdivision means When the size of the circumscribed rectangle is calculated in the vertical direction or the horizontal direction for the circumscribed rectangle determined to be overlapped by the overlap determination in the region extracting device, and any of the sizes is 2 pixels or less And a procedure for determining that the label is a single label having no other overlapping label, and if the determination result is a single label, the coordinate data of the circumscribed rectangle is sent to the output means. Otherwise, a computer-readable recording medium recording a program for realizing the procedure of sending the coordinate data of the circumscribed rectangle to the subdivision means.
[0039]
(Function)
The operation of the present invention based on the above configuration will be described with reference to FIG.
[0040]
(a): In the above (1), the circumscribed rectangle calculating means 12A calculates a circumscribed rectangle of a connected region (connected component of pixels) connected to pixels from the input binary image data, and labels each circumscribed rectangle. Added and stored (for example, stored in a work memory not shown). Next, the overlap determining unit 13A determines whether or not each circumscribed rectangle calculated by the circumscribed rectangle calculating unit 12A overlaps with other circumscribed rectangles.
[0041]
As a result, if there is no overlap, the coordinates of the circumscribed rectangle are sent to the output means 16A, and if there is an overlap, whether or not the single label determination means 14A is a single label for the circumscribed rectangle determined to have overlap. Determine whether. In this case, the single label determination unit 14A calculates the size of the circumscribed rectangle in the vertical direction or the horizontal direction. When the size is 2 pixels or less, the single label determination unit 14A has no single overlapping label inside the label. Judged as one label.
[0042]
If it is determined that the label is a single label, the coordinate data of the circumscribed rectangle is sent to the output means 16A. If it is determined that the label is not a single label, the coordinate data of the circumscribed rectangle is sent to the subdivision means 15A. send.
[0043]
Thereafter, the subdivision means 15A performs a labeling process again to create a label image, and subdivides the circumscribed rectangle of the label image into a parent circumscribing rectangle based on the label contained therein, Repeat the operation to subdivide the child's circumscribed rectangle until the number of labels contained in it is single, determine the coordinates of each child's circumscribed rectangle when the number of labels becomes single, and output the data Send to means.
[0044]
As described above, the circumscribed rectangle determined to be overlapped by the overlap determination unit 13A is subjected to the single label determination unit 14A to determine whether or not it is a single label before performing the subdivision processing by the subdivision unit 15A. If it is determined that the label is a single label, the coordinate data of the corresponding circumscribed rectangle is output to the output means 16A without performing subdivision processing.
[0045]
For this reason, it is not necessary to perform extra subdivision processing for the circumscribed rectangle determined to be a single label, and accordingly, the processing amount can be reduced and the processing speed can be increased. That is, the conventional subdivision means 15A performs a labeling process again on the circumscribed rectangle determined to be overlapped to create a label image, and uses the circumscribed rectangle of the label image as a parent, and labels contained therein Based on this, subdivision into the child's circumscribed rectangle is performed, and further, the subdivision of the child's circumscribed rectangle is repeated until the number of labels contained therein becomes single. It is possible to reduce labeling processing time and the like when performing label determination.
[0046]
This is because the single label determination means 14A calculates the number of labels inside by labeling the number of labels inside by calculating the circumscribing rectangle size when it is known that there are no other labels inside the label. This is because it can be determined without any problem. Therefore, as described above, the processing amount can be reduced and the processing speed can be increased.
[0047]
(b): In (2), when it is determined by the single label determination means 14A that the label is not a single label, the subdivision means 15A performs the labeling process again on the target circumscribed rectangle and performs labeling. An image is created, and the circumscribed rectangle of the label image is used as a parent, and is subdivided into child circumscribed rectangles based on the labels contained therein (the circumscribed rectangle is calculated). The subdivision operation is repeated until the number of contained labels becomes single, and the process of obtaining the coordinates of the circumscribed rectangle of each child when the number of labels becomes single is performed.
[0048]
In this case, the single label determining unit 17A calculates the size of the circumscribed rectangle in the vertical direction or the horizontal direction with respect to the circumscribed rectangle calculated in the process of the subdivision processing, and any one of the sizes is 2 pixels or less. If there is, it is determined that the label is a single label having no other overlapping label. If the determination result is a single label, the subdivision processing for the circumscribed rectangle is terminated, and if it is not a single label, the subdivision processing is further performed for the circumscribed rectangle. In this way, the processing amount of the repeated subdivision process is reduced, and the speed of the subdivision process can be increased.
[0049]
(3): The controller (for example, CPU) of the connected area extraction apparatus reads and executes the program of the recording medium, and thereby determines the overlap rectangle in the connected area extraction process that is determined to have overlap. The vertical or horizontal size of the circumscribed rectangle is calculated, and when one of the sizes is 2 pixels or less, it is determined that the label is a single label having no other overlapping label inside the label. If the determination result is a single label, the circumscribed rectangle coordinate data is sent to the output means, and if it is not a single label, the circumscribed rectangle coordinate data is sent to the subdivision means 15A. To do.
[0050]
As described above, the circumscribed rectangle determined to have overlap in the overlap determination process is determined to be a single label by performing a single label determination process before performing the subdivision process, and determined to be a single label. In this case, the coordinate data of the corresponding circumscribed rectangle is output to the output unit 16A without performing subdivision processing.
[0051]
For this reason, it is not necessary to perform extra subdivision processing, and the processing amount can be reduced and the processing speed can be increased. In other words, conventionally, a circumscribed rectangle determined to have overlap is subjected to a labeling process again to create a label image, and the circumscribed rectangle of the label image is used as a parent, and a child is created based on a label contained therein. Subdivision into circumscribed rectangles and further subdivision of the child's circumscribed rectangle until the number of labels contained therein becomes single is repeated, and single label determination is performed in this subdivision processing. Labeling processing time can be reduced.
[0052]
This is because the single label determination process calculates the number of labels inside by calculating the circumscribing rectangle size when it is known that there are no other labels inside the label. This is because it can be discriminated. Therefore, as described above, the processing amount can be reduced and the processing speed can be increased.
[0053]
DETAILED DESCRIPTION OF THE INVENTION
Embodiments of the present invention will be described in detail below with reference to the drawings.
[0054]
§1: Description of connected region extraction device
FIG. 2 is a block diagram of the connected region extracting apparatus. Hereinafter, based on FIG. 2, a connection area | region extraction apparatus is demonstrated. The connected region extraction device is a device that extracts a connected component (connected region) of pixels from input binary image data and outputs the information as label image information of one screen. The output label image information is: For example, the image is sent to the image processing apparatus and image processing is performed.
[0055]
This connected area extracting device connects pixels from a binary image storage unit 11 (for example, a buffer memory) that temporarily stores input binary image data and the binary image data stored in the binary image storage unit 11. A connected component (or connected region), which is a connected component, is calculated, a circumscribed rectangle of the connected component is calculated, a label is attached to each circumscribed rectangle to form a label image, and this circumscribed rectangle is stored in the work memory When it is determined by the calculation unit 12, the overlap determination unit 13 that determines whether or not each of the calculated circumscribed rectangles overlaps with another circumscribed rectangle, and the overlap determination unit 13 determine that there is an overlap, A single label determination unit 14 that determines whether or not the label is determined, and a subdivision unit 15 that performs further subdivision when the single label determination unit 14 determines that the label is not a single label. rear An output unit 16 for outputting the data provided, further, provided with a single label determination unit 17 to the subdivision unit 15.
[0056]
In this case, the subdivision unit 15 uses the circumscribed rectangle of the label image created by the circumscribed rectangle calculation unit 12 as a parent, and subdivides it into child circumscribed rectangles based on the labels contained therein, and further subdivides the child A process for obtaining the coordinates of the circumscribed rectangle of each child whose number is contained in the circumscribed rectangle is performed. When performing such subdivision processing, the single label determination unit 14 determines whether or not the label is a single label. When the determination is made and it is determined that the label is a single label, the subdivision process is terminated and output to the output unit 16.
[0057]
The other configurations excluding the single label determination units 14 and 17, that is, the configurations of the binary image storage unit 11, circumscribed rectangle calculation unit 12, overlap determination unit 13, subdivision unit 15, and output unit 16 are described above. This is the same as the conventional example (see JP-A-8-55219).
[0058]
§2: Explanation of connected area extraction processing
FIG. 3 is a flowchart of connected area extraction processing. Hereinafter, based on FIG. 3 with reference to FIG. 2, the connected region extraction processing of binary image data will be described. In addition, S1-S8 shows each process step.
[0059]
First, the input binary image data is stored in the binary image storage unit 11 (S1). Thereafter, the circumscribed rectangle calculation unit 12 extracts the binary image data stored in the binary image storage unit 11, and calculates a circumscribed rectangle of a connected component (connected region) in which pixels are connected (S2).
[0060]
In this case, the binary image data is subjected to a labeling process (for example, refer to the specifications and drawings of Japanese Patent Application No. 3-206574 and Japanese Patent Application No. 5-237552), and for each connected pattern (connected component), A circumscribed rectangle of the connected pattern with a label for distinguishing them is obtained and stored in the work memory.
[0061]
Next, the overlap determining unit 13 determines whether or not the circumscribed rectangles of the labels overlap each other with respect to the circumscribed rectangle calculated by the circumscribed rectangle calculating unit 12 (S3). As a result, when it is determined that there is an overlap (S4), the single label determination unit 14 determines whether the label is a single label (S5). However, if it is determined that there is no overlap in the process of S4, the circumscribed rectangle to be determined is output to the output unit 16.
[0062]
Then, it is determined whether or not it is a single label in the process of S6. If it is determined that the label is a single label, the circumscribed rectangle data is output to the output device 16. However, if it is determined in the single label determination process of S5 that the label is not a single label, a circumscribed rectangle is subdivided (details of the subdivision process will be described later) (S7), and the data after subdivision is processed. The output is sent to the output unit 16 (S8). As described above, the connected region extraction process of the binary image data is performed.
[0063]
§3: Explanation of subdivision processing
FIG. 4 is a flowchart of subdivision processing. Hereinafter, the subdivision process will be described with reference to FIG. In addition, S11-S17 show each process step. This process is a process performed by the subdivision unit 15 shown in FIG. 2 and includes a single label determination process by the single label determination unit 17.
[0064]
The subdivision unit 15 performs subdivision processing on circumscribed rectangular data determined by the single label determination unit 14 as not being a single label. In this process, first, the single label determination unit 17 determines whether the single label is a single label. Determine whether or not. That is, it is determined whether or not a label other than the target label (noise label) is a single label (S11).
[0065]
As a result, if it is a single label (S12), the subdivision process is terminated, but if it is not a single label, the subdivision unit 15 performs the circumscribed rectangle calculation process (re circumscribed rectangle calculation process) again (re- circumscribed rectangle calculation process) ( S13). Then, a determination process is performed as to whether or not the calculated circumscribed rectangle is a target label (S14). As a result, if the calculated circumscribed rectangle is not the target label (S15), the subdivision process is terminated, but if it is the target label, the process determines whether the calculated circumscribed rectangle is a single label again ( S16).
[0066]
As a result, if it is determined that the label is a single label (S17), the subdivision process is terminated. If it is determined that the label is not a single label, the process is repeated from S11. As described above, the subdivision processing (S11 to S17) is repeatedly performed until the labels included inside the circumscribed rectangle (noise circumscribed rectangle) other than the target label become single.
[0067]
§4: Description of single label determination processing
FIG. 5 is a flowchart of the single label determination process. FIG. 6 is an explanatory diagram of single label determination processing. FIG. 6A is an overlapping circumscribed rectangle, FIG. 6B is an example 1 of a diagram in which no other overlapping label exists inside the label, and FIG. Figure 2 is an example 2 of a figure in which no label exists, Figure D is an example 1 of a figure in which a label exists, and Figure E is an example 2 of a figure in which a label exists. In FIG. 6, “0” and “1” indicate “white” and “black” pixels of the binary image data.
[0068]
Hereinafter, the single label determination process will be described with reference to FIG. This processing is performed by the single label determination units 14 and 17, and S21 to S27 indicate each processing step. Further, it is assumed that overlapping circumscribed rectangle data and the like are stored in a work memory (not shown).
[0069]
For example, as shown in FIG. 6A, the upper left coordinate point of the overlapping circumscribed rectangle is P1, the lower right coordinate point is P2, and P1 = (x1, y1) and P2 = (x2, y2). And The single label determination units 14 and 17 first obtain the upper left coordinates P1 = (x1, y1) and the lower right coordinates P2 = (x2, y2) of the overlapping circumscribed rectangle from the work memory (S21).
[0070]
Then, the horizontal width (number of pixels in the horizontal direction) W = (x2−x1 + 1) and the height (number of pixels in the vertical direction) H = (y2−y1 + 1) are calculated (S22). Note that the reason why +1 exists in the W and H formulas is that the pixels have a width. Then, the minimum single label size S (preset value S = 2) is acquired (S23), and it is determined whether or not the condition of W ≦ S is satisfied (S24).
[0071]
As a result, if the condition of W ≦ S is satisfied, it is determined as a single label (S26), and the process is terminated. If the condition of W ≦ S is not satisfied, H ≦ S (S = 2) is satisfied. It is determined whether or not the condition is satisfied (S25). As a result, if the condition of H ≦ S is satisfied, it is determined as a single label (S26), and the process is terminated. If the condition of H ≦ S is not satisfied, the conventional process is performed (S27).
[0072]
In the above process, S = 2 (2 pixels), and as shown in FIG. 6B, if W = 2 (horizontal width = 2 pixels), the condition of W ≦ S is satisfied and There are no other overlapping labels. Further, as shown in FIG. 6C, even if W = 2 (width = 2 pixels), if the condition of H ≦ S is satisfied, it is determined that no other overlapping label exists in the label. . Further, when W = 3 (width = 3) as shown in FIG. 6D, or when H = 3 (height = 3) as shown in E, it is determined that a label exists inside. .
[0073]
That is, if W = 2 or less, or H = 2 or less, W or H is 2 pixels or 1 pixel, and therefore, there is no other overlapping label inside the label composed of pixels having such a width. However, if W = 3 or more, or H = 3 or more, there is a possibility that another overlapping label exists inside the label, so the determination is made as described above.
[0074]
As described above, the single label determination units 14 and 17 calculate the horizontal size (horizontal width = W) or the vertical size (height = H) of the circumscribed rectangle to be determined, When the size is 2 pixels or less (W ≦ 2 or H ≦ 2), it is determined that the label is a single label with no other overlapping label inside.
[0075]
In this way, the number of labels inside can be calculated as in the past by calculating the case where it is known that there are no other labels clearly inside the label from the circumscribed rectangle size (W, H). The single label determination process can be performed easily and in a short time without performing the labeling process, and the processing speed can be increased.
[0076]
§5: Description of specific example of connected region extraction device and recording medium
FIG. 7 shows a specific example of the apparatus. The connected area extracting apparatus can be realized by an arbitrary computer such as a personal computer or a workstation. This apparatus includes a computer main body 21, a display device 22 connected to the computer main body 21, an input device (keyboard / mouse, etc.) 23, a removable disk drive (referred to as “RDD”) 24, and a magnetic disk device (referred to as “MDD”). ) 25 etc.
[0077]
The computer main body 21 includes a CPU 26 for performing various internal controls and processes, a ROM 27 for storing programs and various data, a memory 28 (work memory), and an interface control unit (“I / F”). 29) and a communication control unit 30 are provided. The RDD 24 includes a flexible disk drive, an optical disk drive, and the like.
[0078]
In the apparatus having the above configuration, for example, a program for realizing the connected area extraction processing is stored in the ROM 27 or the magnetic disk (recording medium) of the MDD 25, and this program is read and executed by the CPU 26. The connected region extraction process is executed.
[0079]
However, the present invention is not limited to such an example. For example, the program may be stored on the magnetic disk of the MDD 25 as follows, and the CPU 26 may execute the connected region extraction process by executing the program. Is possible.
[0080]
{Circle around (1)} A program (program data created by another device) stored in a removable disk created by another device is read by the RDD 24 and stored in a recording medium of the MDD 25.
[0081]
(2): Data such as a program transmitted from another device via a communication line such as a LAN is received via the communication control unit 30, and the data is stored in a recording medium (magnetic disk) of the MDD 25.
[0082]
【The invention's effect】
As described above, the present invention has the following effects.
[0083]
(1): Since the connected region extraction apparatus is provided with a single label determination means and the subsequent processing is omitted for the circumscribed rectangle determined to be a single label, the processing amount of the entire connected region extraction process is reduced, High speed can be achieved.
[0084]
(2): The single label determination means can calculate the case where it is known that there is no other label inside the label from the circumscribed rectangle size, thereby omitting the labeling process that has been performed conventionally. As a result, it is possible to reduce the processing amount of the connected region extraction processing and to increase the processing speed. In particular, since the blurred image has a lot of fine labels, the above-mentioned effect appears remarkably.
[0085]
(3): For each circumscribed rectangle calculated by the circumscribed rectangle calculating means, the overlap determining means determines whether or not there is an overlap with other circumscribed rectangles. If there is no overlap, the coordinates of the circumscribed rectangle are sent to the output means, and the overlap is determined. If there is, the single label determination means determines whether or not the circumscribed rectangle determined to have overlap is a single label.
[0086]
In this case, the single label determination means calculates the size of the circumscribed rectangle in the vertical or horizontal direction, and when the size is 2 pixels or less, the single label has no other overlapping label inside the label. It is determined that If it is determined that the label is a single label, the coordinate data of the circumscribed rectangle is sent to the output means. If it is determined that the label is not a single label, the coordinate data of the circumscribed rectangle is sent to the subdivision means.
[0087]
In this way, for the circumscribed rectangle determined to be overlapped by the overlap determination unit, before performing the subdivision process by the subdivision unit, it is determined whether or not it is a single label by the single label determination process. Is determined, the corresponding circumscribed rectangle coordinate data is output to the output means without performing subdivision processing.
[0088]
For this reason, it is not necessary to perform the subdivision processing for the circumscribed rectangle determined to be a single label, and accordingly, the processing amount can be reduced and the processing speed can be increased. That is, in the conventional subdivision means, a label image is created again by performing a labeling process on the circumscribed rectangle determined to be overlapped, and the label contained within the circumscribed rectangle is used as a parent. Then, subdividing the child's circumscribing rectangle and further subdividing the child's circumscribing rectangle until the number of labels contained in the child becomes single, but the circumscribing rectangle determined to be a single label This subdivision process can be reduced.
[0089]
In addition, by calculating the case where it is determined that there is no other label inside the label from the circumscribed rectangle size, the labeling process that has been performed conventionally can be omitted. As a result, it is possible to reduce the processing amount of the connected region extraction processing and to increase the processing speed. In particular, since the blurred image has a lot of fine labels, the above-mentioned effect appears remarkably. Therefore, it is possible to reduce the processing amount of the connected area extraction process and achieve a high-speed process.
[0090]
(4): The circumscribing rectangle determined to be overlapped with other circumscribed rectangles by the overlap determining means is determined as to whether it is a single label, and the circumscribing rectangle determined not to be a single label is subdivided. I do. In this subdivision processing, the subdivision means performs labeling processing again on the corresponding circumscribed rectangle, creates a label image, and uses the circumscribed rectangle of the label image as a parent, based on the label contained therein, Subdividing the child's circumscribed rectangle into sub-boundary rectangles and further subdividing the child's circumscribed rectangle until the number of labels contained in the child becomes single. Performs processing to obtain rectangular coordinates.
[0091]
In this case, the single label determination unit in the subdivision unit calculates the vertical or horizontal size of the circumscribed rectangle for the circumscribed rectangle calculated in the process of the subdivision process, and any one of the sizes is If it is 2 pixels or less, it is determined that the label is a single label with no other overlapping label inside. If the determination result is a single label, the subdivision process for the circumscribed rectangle is terminated.
[0092]
In this way, when the subdivision processing is performed, the subsequent processing can be omitted for the circumscribed rectangle determined to be a single label. Accordingly, the processing amount of the repeated subdivision process is reduced, and the speed of the subdivision process can be increased. In this case, in particular, the labeling process or the like can be omitted by calculating from the circumscribed rectangle size the case where it is determined that there is no other label inside the label.
[0093]
As a result, in particular, the processing amount can be increased by reducing the processing amount of the subdivision processing. In particular, since the blurred image has many fine labels, the above-described effect appears remarkably when the blurred image is handled.
[0094]
(5): The controller (for example, CPU) of the connected area extraction apparatus reads and executes the program of the recording medium, thereby determining the overlap rectangle in the connected area extraction process that is determined to have overlap. The vertical or horizontal size of the circumscribed rectangle is calculated, and when one of the sizes is 2 pixels or less, it is determined that the label is a single label having no other overlapping label inside the label. If the determination result is a single label, the circumscribed rectangle coordinate data is sent to the output means, and if it is not a single label, the circumscribed rectangle coordinate data is sent to the subdivision means. .
[0095]
As described above, the circumscribed rectangle determined to have overlap in the overlap determination process is determined to be a single label by performing a single label determination process before performing the subdivision process, and determined to be a single label. In this case, the coordinate data of the corresponding circumscribed rectangle is output to the output means without performing subdivision processing.
[0096]
For this reason, it is not necessary to perform extra subdivision processing, and the processing amount can be reduced and the processing speed can be increased. In other words, conventionally, a circumscribed rectangle determined to have overlap is subjected to a labeling process again to create a label image, and the circumscribed rectangle of the label image is used as a parent, and a child is created based on a label contained therein. Subdivision into circumscribed rectangles and further subdivision of the circumscribed rectangles of its children until the number of labels contained therein becomes single are repeated, but this subdivision processing can be reduced.
[0097]
In addition, by calculating the case where it is determined that there is no other label inside the label from the circumscribed rectangle size, the labeling process that has been conventionally performed can be omitted. As a result, it is possible to reduce the processing amount of the connected region extraction processing and to increase the processing speed. In particular, since the blurred image has a lot of fine labels, the above-mentioned effect appears remarkably. Therefore, it is possible to reduce the processing amount of the connected area extraction process and achieve a high-speed process.
[Brief description of the drawings]
FIG. 1 is a diagram illustrating the principle of the present invention.
FIG. 2 is a block diagram of a connected region extraction apparatus in an embodiment of the present invention.
FIG. 3 is a flowchart of connected area extraction processing in the embodiment of the present invention.
FIG. 4 is a flowchart of subdivision processing in the embodiment of the present invention.
FIG. 5 is a flowchart of a single label determination process in the embodiment of the present invention.
FIG. 6 is an explanatory diagram of single label determination processing according to the embodiment of the present invention.
FIG. 7 is a specific apparatus example according to the embodiment of the present invention.
FIG. 8 is an explanatory diagram (part 1) of a conventional example.
FIG. 9 is an explanatory diagram (part 2) of a conventional example.
FIG. 10 is an explanatory diagram of a conventional example (part 3);
[Explanation of symbols]
1 connected component extraction unit
2 Connected component subdivision
3 Subdivision area determination unit
4 Subdivision area storage
11 Binary image storage unit
12 circumscribed rectangle calculator
13 Overlap determination unit
14 Single label judgment part
15 Subdivision section
16 Output section
17 Single label determination unit
21 Computer body
22 Display device
23 Input device
24 Removable disk drive
25 Magnetic disk unit
26 CPU (Central Processing Unit)
27 ROM (Read Only Memory)
28 memory
29 Interface control unit (I / F control unit)
30 Communication control unit

Claims (3)

Circumscribing rectangle calculation means for calculating a circumscribed rectangle of a connected area where pixels are connected from the input binary image data, and storing each circumscribed rectangle with a label;
For each circumscribed rectangle calculated by the circumscribed rectangle calculating means, an overlap determining means for determining whether or not there is an overlap with other circumscribed rectangles;
For the circumscribed rectangle determined to be overlapped by the determination of the overlap determining means, a label image is created again by creating a label image, based on the label contained inside the circumscribed rectangle of the label image as a parent, Subdivide the child's circumscribing rectangle into sub-divisions, and further subdivide the child's circumscribing rectangle until the number of labels contained in it is unity. Subdivision means for obtaining the coordinates of the circumscribed rectangle;
Data consisting of the input binary image data, the coordinates of the circumscribed rectangle determined to be non-overlapping by the determination of the overlap determining means, and the circumscribed rectangle of the parent and each child determined by the subdivision means In the connected region extraction apparatus having output means for outputting,
For the circumscribed rectangle determined to have overlap by the determination of the overlap determining means, the vertical or horizontal size of the circumscribed rectangle is calculated, and when any of the sizes is 2 pixels or less, A single label determination means for determining that there is no other overlapping label inside the label, and
If the determination result is a single label, the coordinate data of the circumscribed rectangle is sent to the output means, and if it is not a single label, the coordinate data of the circumscribed rectangle is sent to the subdivision means.
A connected region extracting apparatus. The subdivision means includes
For the circumscribed rectangle calculated in the process of performing the subdivision processing of the circumscribed rectangle, the size of the circumscribed rectangle is calculated in the vertical direction or the horizontal direction. A single label determination means for determining that there is no other overlapping label inside the label, and
If the determination result of the single label determination means is a single label, the subdivision processing for the circumscribed rectangle is terminated, and if it is not a single label, the subdivision processing for the circumscribed rectangle is performed. ing,
The connected area extracting apparatus according to claim 1, wherein: Circumscribed rectangle calculating means for calculating a circumscribed rectangle of a connected region where pixels are connected from input binary image data, and storing each circumscribed rectangle with a label;
For each circumscribed rectangle calculated by the circumscribed rectangle calculating means, an overlap determining means for determining whether or not there is an overlap with other circumscribed rectangles;
For the circumscribed rectangle determined to be overlapped by the determination of the overlap determining means, a labeling process is performed again to create a label image, with the circumscribed rectangle of the label image as a parent, based on the label contained therein, Subdividing the child's circumscribed rectangle into sub-boundaries and further subdividing the child's circumscribed rectangle until the number of labels contained in the child becomes single. Subdivision means for obtaining rectangular coordinates;
Data consisting of the input binary image data, the coordinates of the circumscribed rectangle determined to be non-overlapping by the determination of the overlap determining means, and the circumscribed rectangle of the parent and each child determined by the subdivision means In the connected region extraction apparatus provided with output means for outputting,
For the circumscribed rectangle determined to be overlapped by the overlap determination, the vertical or horizontal size of the circumscribed rectangle is calculated, and if any of the sizes is 2 pixels or less, the label inside A procedure for determining that a single label has no other overlapping label,
If the determination result is a single label, the coordinate data of the circumscribed rectangle is sent to the output means, and if it is not a single label, the coordinate data of the circumscribed rectangle is sent to the subdivision means;
The computer-readable recording medium which recorded the program for implement | achieving.

Images