JP4611224B2 - Image feature area dividing program and computer-readable recording medium - Google Patents

Description

The present invention relates to an image feature area dividing program and a computer-readable recording medium .

The process of dividing an area from within an image is a process that is the basis of image processing. An example of this image region division processing is disclosed in Patent Document 1. The region segmentation method disclosed in Patent Document 1 is performed by calculating a feature amount of an input image using a Sobel filter or the like and analyzing the feature amount.
JP-A-7-296171

  However, information of at least two pixels or more is required for calculating the feature amount indicating the degree of edge or line. For this reason, in the conventional method, only wide-area information can be obtained, and an accurate edge position cannot be detected. In particular, the feature amount of a slight flat region between edges such as between characters is a value that cannot be distinguished from the feature amount of the edge region due to surrounding edges.

The present invention has been made in order to solve the above problems, an object of the image characteristic region dividing program can be represented as a small flat area like Accurately feature amount between the object and the object and which It is to provide a recorded computer-readable recording medium .

Another object of the present invention is to provide an image feature area dividing program that is excellent in resolution of feature amounts, and can finely and accurately divide a target feature area from other areas, and a computer-readable recording medium on which the image feature area is recorded. It is to be.

An image feature region dividing program according to an aspect of the present invention includes an enlargement processing step for enlarging an input image including an object region in which a plurality of objects are arranged at predetermined intervals, and a feature of the image enlarged by the enlargement processing step. A feature amount calculation processing step for calculating a feature amount for each pixel of the image using a plurality of computation operators for the pixel and a predetermined range of pixels in the vicinity thereof, and the feature amount A feature amount classification processing step for classifying each pixel into a feature region and a non-feature region by thresholding each pixel of the feature amount data created in the calculation processing step; The feature areas of the classification data created by the feature quantity classification processing step are arranged at the predetermined intervals. A first region expansion processing step for expanding a region by a first expansion amount for integrating at least a part of the feature region without integrating each object in the object region; and the first region Among the small pieces constituting the feature region expanded by the expansion processing step, a small piece having a predetermined first size or more exceeding the size of the object arranged at the predetermined interval is deleted from the feature region. The feature areas in which the pieces larger than the first size are deleted by the first area deletion processing step and the first area deletion processing step are integrated with each object in the object area arranged at the predetermined interval. A second region expansion processing step for expanding the region with a second expansion amount for causing each of the characteristic regions From the feature area, a small piece having a size equal to or smaller than a predetermined second size less than the size of the object area arranged at the predetermined interval integrated by the second area expansion processing step is deleted from the feature area. And a second area deletion processing step.

An image feature region dividing program according to another aspect of the present invention includes an enlargement processing step for enlarging an input image including an object region in which a plurality of objects are arranged at predetermined intervals, and an image enlarged by the enlargement processing step. A feature amount calculation processing step for calculating a feature amount for each pixel of the image using a plurality of calculation operators for the pixel and pixels in a predetermined range in the vicinity thereof, and the feature A feature amount classification processing step for classifying each pixel into a feature region and a non-feature region by thresholding each pixel of the feature amount data created in the amount calculation processing step and creating classification data And the feature areas of the classification data created by the feature quantity classification processing step are arranged at the predetermined intervals. A first region expansion processing step for expanding a region by a first expansion amount for integrating at least a part of the feature region without integrating each object in the object region; and the first region Among the small pieces constituting the feature region expanded by the expansion processing step, a small piece having a predetermined first size or more exceeding the size of the object arranged at the predetermined interval is deleted from the feature region. The feature areas in which the pieces larger than the first size are deleted by the first area deletion processing step and the first area deletion processing step are integrated with each object in the object area arranged at the predetermined interval. A second region expansion processing step for expanding the region with a second expansion amount for causing each of the characteristic regions From the feature area, a small piece having a size equal to or smaller than a predetermined second size less than the size of the object area arranged at the predetermined interval integrated by the second area expansion processing step is deleted from the feature area. A second region deletion processing step for reducing the size of the classification data, and a reduction processing step for reducing the size of the classification data.

An image feature area dividing program according to another aspect of the present invention includes an enlargement processing step for enlarging an input image including an object area in which a plurality of objects are arranged at predetermined intervals;
  In order to create feature amount data by calculating the feature amount of the image enlarged by the enlargement processing step for each pixel of the image using a plurality of arithmetic operators for the pixel and a predetermined range of pixels in the vicinity thereof. The feature amount calculation processing step, a reduction processing step for reducing the size of the feature amount data created by the feature amount calculation processing step, and each pixel of the feature amount data reduced by the reduction processing step. By performing threshold processing, each pixel is classified into a feature region and a non-feature region, a feature amount classification processing step for creating classification data, and the classification data generated by the feature amount classification processing step The feature area of the object areas may not be integrated with each other in the object areas arranged at the predetermined intervals. A first region expansion processing step for expanding the region by a first expansion amount for integrating at least a part of the feature region, and the feature region expanded by the first region expansion processing step are configured. A first area deletion processing step for deleting, from the feature area, a small piece having a predetermined first size or more that exceeds the size of the object arranged at the predetermined interval, and the first area In order to expand the feature region in which the pieces larger than the first size are deleted by the deletion processing step by the second expansion amount for integrating the objects of the object regions arranged at the predetermined interval. The second region expansion processing step and the second region expansion processing step out of the small pieces constituting the feature region. And and a second region removing processing steps for a second size less than the size of pieces determined in advance under the predetermined interval size of the deployed object area to remove from the feature region.

  The resolution of the feature amount is excellent, and the target feature area can be divided from other areas finely and accurately.

  Further, by performing expansion and contraction of the feature region in stages, it is possible to narrow down the target feature region determination conditions.

  Further, after the input image is enlarged, the feature amount is calculated. Therefore, by using this feature amount, an object in a complex image can be extracted. Also, the position of the extracted object is accurate.

[First Embodiment]
Hereinafter, the feature amount data used to describe the embodiment is defined as numerical data representing an arbitrary feature such as an output value of a Sobel filter used for detecting an edge from an image.

  With reference to FIG. 1, an image feature area dividing device 102 according to a first embodiment of the present invention will be described.

  The image feature region dividing device 102 receives the input image data from the image input device 101 and outputs the region divided data to the region divided data output device 103.

  The image feature area dividing device 102 is connected to the input image data buffer 104 for storing the input image data input from the image input device 101 and the input image data buffer 104, and reads the input image data from the input image data buffer 104. A feature amount extraction processing unit 105 that extracts feature amounts and outputs feature amount data, and a feature amount data buffer that is connected to the feature amount extraction processing unit 105 and stores feature amount data output from the feature amount extraction processing unit 105 106 is connected to the feature amount data buffer 106, reads the feature amount data from the feature amount data buffer 106, performs region extraction based on the feature amount data, and outputs the region division data to the region division data output device 103. Part 107.

  The input image data buffer 104 and the feature data buffer 106 are realized by a RAM (Random Access Memory) such as a flash memory or a hard disk. The feature amount extraction processing unit 105 and the region extraction processing unit 107 are realized by independent circuits. The feature amount extraction processing unit 105 and the region extraction processing unit 107 may be virtual circuits realized by an arithmetic processing circuit such as a computer.

  Referring to FIG. 2, the feature amount extraction processing unit 105 shown in FIG. 1 is connected to the input image data buffer 104, reads data from the input image data buffer 104, enlarges the input image, and outputs enlarged image data. An enlargement processing unit 201, an enlargement image data buffer 202 that is connected to the enlargement processing unit 201 and stores enlarged image data output from the enlargement processing unit 201, and an enlargement image data buffer 202 that is connected to the enlargement image data buffer 202. And a feature amount calculation processing unit 203 that reads an enlarged image, calculates a feature amount, and outputs feature amount data.

  The enlarged image data buffer 202 is realized by a RAM such as a flash memory or a hard disk. The enlargement processing unit 201 and the feature amount calculation processing unit 203 are realized by independent circuits. The enlargement processing unit 201 and the feature amount calculation processing unit 203 may be a virtual circuit realized by an arithmetic processing circuit such as a computer.

  Referring to FIG. 3, region extraction processing unit 107 in FIG. 1 is connected to feature amount data buffer 106, reads feature amount data from feature amount data buffer 106, performs threshold processing on feature amount data, And a feature quantity classification processing unit 301 that outputs classification data composed of non-feature regions, and a feature quantity classification processing unit 301 and a region deletion processing unit 304 (described later), and the feature quantity classification processing unit 301 or the region deletion processing unit 304 Is connected to the classification data buffer 302 for storing the classification data output from, the classification data buffer 302, the region expansion processing unit 306 (described later) and the data reduction processing unit 305 (described later), and reads the classification data from the classification data buffer 302 , Classification data determination processing to be output to either the region expansion processing unit 306 or the data reduction processing unit 305 303 and a region expansion processing unit 306 connected to the classification data determination determination processing unit 303, receiving the classification data from the classification data determination determination processing unit 303, expanding the feature region by a predetermined pixel, and outputting the region expansion data. Including.

  The region extraction processing unit 107 is further connected to the region expansion processing unit 306, stores a region expansion data buffer 307 for storing the region expansion data output from the region expansion processing unit 306, and receives the region expansion data from the region expansion data buffer 307. An area reduction processing unit 308 that reads and reduces the feature area by a predetermined pixel and outputs area reduction data, and an area that is connected to the area reduction processing unit 308 and stores area reduction data output from the area reduction processing unit 308 Connected to the reduced data buffer 309, the area reduced data buffer 309, and the classification data buffer 302, reads the area reduced data from the area reduced data buffer 309, deletes a small area based on the area reduced data, and then reduces the area Data is stored in the classification data buffer 302 as classification data. A data reduction process that is connected to the region deletion processing unit 304 and the classification data determination determination processing unit 303, receives the classification data from the classification data determination determination processing unit 303, reduces the classification data, and then outputs the classification data as region division data Part 305.

  The classification data buffer 302, the area expansion data buffer 307, and the area degeneration data buffer 309 are realized by a RAM such as a flash memory or a hard disk.

  The feature amount classification processing unit 301, the classification data determination processing unit 303, the region expansion processing unit 306, the region reduction processing unit 308, the region deletion processing unit 304, and the data reduction processing unit 305 are realized by independent circuits. The feature amount classification processing unit 301, the classification data determination processing unit 303, the region expansion processing unit 306, the region reduction processing unit 308, the region deletion processing unit 304, and the data reduction processing unit 305 are realized by an arithmetic processing circuit such as a computer. It may be a virtual circuit.

  With reference to FIG. 4, the process of the image feature area dividing device 102 according to the first embodiment will be described. Here, a case where a character region of an image as shown in FIG. 5 is extracted is used as a specific example.

  The image input apparatus 101 stores the input image in the input image data buffer 104 (step S402; hereinafter, “step” is omitted).

  The enlargement processing unit 201 enlarges the input image at a magnification of 3 times using the 0th-order hold method (S403). The zero-order hold method is a well-known technique. Therefore, detailed description thereof will not be repeated here. Any magnification may be used as long as it is 1 or more. Here, the 0th-order hold method is used as the enlargement method, but any enlargement method may be used.

  The enlargement processing unit 201 stores the enlarged image data in the enlarged image data buffer 202 (S404).

  The feature amount calculation processing unit 203 calculates feature amounts such as line drawings, edges, and screen tones from the enlarged image data (S405).

  Here, for each pixel, four feature amounts are calculated using the four arithmetic operators shown in FIGS. 6A to 6D for the pixel and its eight neighboring pixels, and the maximum value is calculated. The product of 255/2296 is the feature quantity of the pixel. Here, feature data as shown in FIG. 7 is created.

  Note that the value calculated by the arithmetic operator is multiplied by 255/2296 in order to keep the feature amount in the range of 0 to 255, but the range may be any range. Any method such as a Sobel filter may be used to calculate the feature amount.

  The feature amount calculation processing unit 203 stores the feature amount data in the feature amount data buffer 106 (S406).

  The feature quantity classification processing unit 301 classifies the feature quantity data into a feature area and a non-feature area using the value of the feature quantity data (S407). Here, an area having a feature value of 100 or more using a threshold value 100 is defined as a feature area, and an area having a feature value less than 100 is defined as a non-feature area. As a result, classification data as shown in FIG. 8 is created.

  The feature amount classification processing unit 301 stores the classification data in the classification data buffer 302 (S408).

  The classification data determination processing unit 303 determines whether or not the number of times classification data is written to the classification data buffer 302 has reached three (S409).

  If the classification data has been written three times (YES in S409), the classification data determination processing unit 303 regards that the classification data in the classification data buffer 302 has been determined, and the classification data is sent to the data reduction processing unit 305. And the process of S415 is executed. If the number of times of writing the classification data is less than 3 (NO in S409), the classification data in the classification data buffer 302 is passed to the area expansion processing unit 306, and the processes after S410 are executed.

  Note that the operation switching reference of the classification data determination processing unit 303 does not need to be written three times in the classification data buffer 302 and can be freely set.

  Now, the classification data is written into the classification data buffer 302 once. For this reason, the region expansion processing unit 306 receives the classification data from the classification data decision determination processing unit 304, expands the feature region by 2 pixels, and creates region expansion data as shown in FIG. 9 (S410). Here, the reason why the pixels are expanded by two pixels is to integrate the feature regions that have been cut into pieces.

  The region expansion processing unit 306 stores the region expansion data in the region expansion data buffer 307 (S411).

  The area reduction processing unit 308 reads the area expansion data from the area expansion data buffer 307 and creates area reduction data in which the feature area is reduced by 0 pixels (S412). That is, in this example, the region degeneration process is not substantially performed.

  The area reduction processing unit 308 stores the area reduction data in the area reduction data buffer 309 (S413).

  The area deletion processing unit 304 counts the number of pixels in each small piece of the feature area by four connections, and creates classification data as shown in FIG. 10 excluding small pieces having a pixel number of 1800 pixels or more from the feature area (S414). ). The created classification data is written into the classification data buffer 302 (S408).

  Since characters are captured by humans as objects, they are usually spaced one character at a time. In this process (S414), it is determined that a small piece made of a pixel having an unnatural size as a character size in the feature area is not a character.

  Next, when the determination in S409 is performed, the number of times classification data is written to the classification data buffer 302 is two. For this reason, the process after S410 is performed again.

  The region expansion processing unit 306 receives the classification data as shown in FIG. 10 from the classification data decision determination processing unit 303, creates region expansion data in which the feature region is expanded by 30 pixels, and writes it into the region expansion data buffer 307. (S410, S411).

  The area reduction processing unit 308 reads the area expansion data from the area expansion data buffer 307, reduces the feature area by 30 pixels, and generates area reduction data as shown in FIG. 11 (S412). The character string becomes a small piece of a certain size by the second expansion / contraction. The number of expanded pixels and the number of degenerated pixels depending on the number of repetitions can be freely set.

  The area reduction processing unit 308 stores the area reduction data in the area reduction data buffer 309 (S413).

  The area deletion processing unit 304 counts by connecting four pixels in each small piece of the feature area, removes small pieces having a pixel number of 900 pixels or less from the feature area, and creates classification data as shown in FIG. 12 (S414). The created classification data is written into the classification data buffer 302 (S408).

  The condition of the small piece to be deleted can be freely set according to the number of repetitions. The count of the number of pixels in the small piece may be counted by 8 connections.

  Here, since the number of times the classification data is written to the classification data buffer 302 has reached three (YES in S409), the classification data determination processing unit 303 passes the classification data to the data reduction processing unit 305.

  The data reduction processing unit 305 reduces the classification data to 1/3 and outputs area division data as shown in FIG. 13 to the area division data output device 103. The reduction process is performed according to the reciprocal of the enlargement magnification in the enlargement process (S415).

  The reduction process converts 3 × 3 pixel data into 1 pixel data. If there is a 3 × 3 pixel data designated as a feature area, the converted 1 pixel data is used as a feature area. Specify the others, and make them non-characteristic areas. Note that this reduction process may be any method.

  As described above, according to the present embodiment, after the input image is enlarged, the feature amount is obtained, and the feature region is extracted based on the feature amount. Thus, since the resolution of the feature amount is excellent, the target feature region can be divided from other regions finely and accurately.

  Further, the expansion and contraction of the region are performed in stages. For this reason, it is possible to narrow down the conditions for determining the target feature region.

[Second Embodiment]
Referring to FIG. 14, image feature region dividing device 110 according to the second exemplary embodiment of the present invention is configured as image feature region dividing device 102 according to the first exemplary embodiment described with reference to FIG. 1. 1, the feature amount extraction processing unit 108 is used instead of the feature amount extraction processing unit 105, and the region extraction processing unit 109 is used instead of the region extraction processing unit 107. Other components are the same as those in the first embodiment. Therefore, detailed description thereof will not be repeated here.

  The feature amount extraction processing unit 108 reads input image data from the input image data buffer 104, extracts the feature amount, and outputs the feature amount data. The region extraction processing unit 109 reads the feature amount data from the feature amount data buffer 106, extracts a region based on the feature amount data, and outputs the region division data to the region division data output device 103.

  The feature amount extraction processing unit 108 and the region extraction processing unit 109 are realized by independent circuits. The feature amount extraction processing unit 108 and the region extraction processing unit 109 may be virtual circuits realized by an arithmetic processing circuit such as a computer.

  Referring to FIG. 15, the feature amount extraction processing unit 108 includes an enlargement processing unit 201 connected to the input image data buffer 104, an enlarged image data buffer 202 connected to the enlargement processing unit 201, and an enlarged image data buffer 202. Connected to the feature amount calculation processing unit 203, a reduction amount processing unit 604 connected to the feature amount calculation processing unit 203, which reduces the feature amount data and outputs the reduced feature amount data to the feature amount data buffer 106, and including.

  The enlargement processing unit 201, the enlarged image data buffer 202, and the feature amount calculation processing unit 203 are the same as those described with reference to FIG. Therefore, detailed description thereof will not be repeated here.

  The reduction processing unit 604 is realized by an independent circuit. The reduction processing unit 604 may be a virtual circuit realized by an arithmetic processing circuit such as a computer.

  Referring to FIG. 16, the region extraction processing unit 109 in FIG. 14 is obtained by removing the data reduction processing unit 305 from the configuration of the region extraction processing unit 107 described with reference to FIG. For this reason, description of each component is not repeated here. The output of the classification data determination processing unit 303 is connected to the area division data output device 103 instead of being connected to the data reduction processing unit 305.

  With reference to FIG. 17, the process of the image feature region dividing device 110 according to the second embodiment will be described. Here, a case where a character region of an image as shown in FIG. 5 is extracted is used as a specific example.

  The processing from S802 to S805 is the same as the processing from S402 to S405 in FIG. Therefore, detailed description thereof will not be repeated here.

  The reduction processing unit 604 reduces the feature amount data to 1/3 (S806). In the reduction method, the value of the maximum pixel value among the original nine pixels that become one pixel after reduction is used as the pixel value after reduction. Although this reduction method has an effect of preventing the formation of a cut feature region after reduction, other reduction methods such as an average value operation method may be used. The reduction process is performed according to the inverse of the enlargement magnification of the enlargement process. The reduced feature data is stored in the feature data buffer 106 (S807).

  The feature quantity classification processing unit 301 classifies the feature quantity data into a feature area and a non-feature area using the value of the feature quantity data (S808). Here, an area having a feature value of 100 or more using a threshold value 100 is defined as a feature area, and an area having a feature value less than 100 is defined as a non-feature area. As a result, classification data as shown in FIG. 18 is created.

The feature amount classification processing unit 301 stores the classification data in the classification data buffer 302.
The classification data determination processing unit 303 determines whether or not the number of times classification data is written to the classification data buffer 302 has reached three (S810).

  If the classification data has been written three times (YES in step S810), the classification data determination processing unit 303 regards that the classification data in the classification data buffer 302 has been determined, and classifies the classification data into the region division data output device 103. Data is output and the process ends. If the number of times of writing the classification data is less than 3 (NO in S810), the classification data in the classification data buffer 302 is passed to the area expansion processing unit 306, and the processes after S811 are executed.

  Note that the operation switching reference of the classification data determination processing unit 303 does not need to be written three times in the classification data buffer 302 and can be freely set.

  Now, the classification data is written into the classification data buffer 302 once. Therefore, the region expansion processing unit 306 receives the classification data from the classification data determination determination processing unit 303, expands the feature region by 0 pixels, and creates region expansion data (S812). That is, in this example, the area expansion process is not substantially performed here.

  The region expansion processing unit 306 stores the region expansion data in the region expansion data buffer 307 (S812).

  The area reduction processing unit 308 reads the area expansion data from the area expansion data buffer 307 and creates area reduction data in which the feature area is reduced by 0 pixels (S813). That is, in this example, the region degeneration process is not substantially performed.

  The area reduction processing unit 308 stores the area reduction data in the area reduction data buffer 309 (S814).

  The area deletion processing unit 304 counts the number of pixels of the feature area pieces in four concatenations, and creates classification data as shown in FIG. 19 in which the pieces having a pixel number of 200 pixels or more are excluded from the feature area (S815).

The created classification data is written into the classification data buffer 302 (S809).
Since characters are captured by humans as objects, they are usually spaced one character at a time. In this process (S815), it is determined that a small piece made of a pixel having an unnatural size as a character size in the feature area is not a character.

  Next, when the determination in S810 is performed, the number of times classification data is written to the classification data buffer 302 is two. For this reason, the processing after S811 is executed again.

  The region expansion processing unit 306 receives the classification data from the classification data determination / determination processing unit 303, creates region expansion data in which the feature region is expanded by 10 pixels, and writes it into the region expansion data (S811, S812).

  The area reduction processing unit 308 reads the area expansion data from the area expansion data buffer 307, reduces the feature area by 10 pixels, and creates area reduction data as shown in FIG. 20 (S813). The character string becomes a small piece of a certain size by the second expansion / contraction.

  The area reduction processing unit 308 stores the area reduction data in the area reduction data buffer 309 (S814).

  The region deletion processing unit 304 counts the feature region pieces by counting the number of pixels in a four-connected manner, removes the piece having a pixel number of 100 pixels or less from the feature region, and creates data as shown in FIG. 21 as classification data (S815). The created classification data is written into the classification data buffer 302 (S809).

  The condition of the small piece to be deleted can be freely set according to the number of repetitions. The count of the number of pixels in the small piece may be counted by 8 connections.

  Here, since the number of times the classification data is written to the classification data buffer 302 has reached three (YES in S810), the classification data is output to the area division data output device 103 as area division data.

  As described above, according to the present embodiment, after the input image is enlarged, the feature amount is obtained, and the feature region is extracted based on the feature amount. Thus, since the resolution of the feature amount is excellent, the target feature region can be divided from other regions finely and accurately.

  Further, the expansion and contraction of the region are performed in stages. For this reason, it is possible to narrow down the conditions for determining the target feature region.

  Furthermore, the feature region is extracted based on feature amount data having the same size as the input image. For this reason, compared with the first embodiment, there is less execution processing of the region extraction processing unit, and high-speed processing is possible. Further, the capacity of the buffer for storing the classification data can be reduced.

[Third Embodiment]
With reference to FIG. 22, an image feature amount calculation apparatus 902 according to a third embodiment of the present invention will be described.

  The image feature amount calculation device 902 receives an input image from the image input device 101 and outputs feature amount data to the feature amount data output device 903.

  The image feature amount calculation device 902 is connected to the image input device 101, and is connected to the input image data buffer 104 that stores the image data input from the image input device 101 and the input image data buffer 104. An enlargement processing unit 201 that reads data from and outputs enlarged image data, an enlarged image data buffer 202 that is connected to the enlargement processing unit 201 and stores enlarged image data output from the enlargement processing unit 201, and an enlarged image data buffer A feature amount calculation processing unit 203 that is connected to 202 and reads the enlarged image from the enlarged image data buffer 202 and outputs the feature amount data to the feature amount data output device 903.

  The input image data buffer 104 and the enlarged image data buffer 202 are realized by a RAM such as a flash memory or a hard disk. The enlargement processing unit 201 and the feature amount calculation processing unit 203 are realized by independent circuits. The enlargement processing unit 201 and the feature amount calculation processing unit 203 may be a virtual circuit realized by an arithmetic processing circuit such as a computer.

  With reference to FIG. 23, the process of the image feature amount calculation apparatus 902 according to the third embodiment will be described.

  An input image from the image input device 101 is stored in the input image data buffer 104 (S1002).

  The enlargement processing unit 201 enlarges the input image at a magnification of 3 times using the 0th-order hold method (S1003). Any magnification may be used as long as it is 1 or more. Here, the 0th-order hold method is used as the enlargement method, but any enlargement method may be used.

  The enlargement processing unit 201 stores the enlarged image data in the enlarged image data buffer 202 (S1004).

  The feature amount calculation processing unit 203 calculates feature amounts such as line drawings, edges, and screen tones from the enlarged image data (S1005).

  Here, for each pixel, the feature amount is calculated using the four arithmetic operators shown in FIGS. 6A to 6D for the pixel and the eight neighboring pixels, and the maximum value is calculated as the maximum value. The feature amount of the pixel.

Any method such as a Sobel filter may be used to calculate the feature amount.
The feature amount calculation processing unit 203 outputs the feature amount data to the feature amount data output device 903 (S1006).

  As described above, according to the present embodiment, the feature amount is calculated after enlarging the input image. Therefore, by using this feature amount, an object in a complex image can be extracted. Also, the position of the extracted object is accurate.

  In addition, since the image feature quantity calculation device outputs high-resolution data, it is possible to analyze a detailed feature quantity from the output result.

[Fourth Embodiment]
With reference to FIG. 24, an image feature quantity calculation apparatus 1102 according to the fourth embodiment of the present invention will be described.

  The image feature amount calculation device 1102 receives an input image from the image input device 101 and outputs feature amount data to the feature amount data output device 903.

The image feature amount calculation apparatus 1102 is obtained by adding a reduction processing unit 604 to the hardware configuration of the image feature amount calculation apparatus 902 according to the third embodiment. The reduction processing unit 604 is the same as that described with reference to FIG. The reduction processing unit 604 outputs the feature amount data obtained by reducing the feature amount data output from the feature amount calculation processing unit 203 to the feature amount data output device 903. Other components are the same as those in the third embodiment. Therefore, detailed description thereof will not be repeated here. With reference to FIG. 25, processing of the image feature amount calculation apparatus 1102 according to the fourth embodiment will be described.

  The processing of S1202 to S1205 is the same as the processing of S1002 to S1005 described with reference to FIG. Therefore, detailed description thereof will not be repeated here.

  In the reduction processing unit 604, the feature amount data is reduced to ３ using the average value operation method and is output to the feature amount data output device 903 (S1206). The average value manipulation method is a well-known technique. Therefore, detailed description thereof will not be repeated here. The reduction process is performed according to the reciprocal of the enlargement magnification in the enlargement process. Note that the reduction processing method may be a method other than the average value operation method. Here, the reduction process is performed in order to reduce the size of the feature data.

  As described above, according to the present embodiment, the feature amount is calculated after enlarging the input image. Therefore, by using this feature amount, an object in a complex image can be extracted. Also, the position of the extracted object is accurate.

In addition, the resolution of the feature amount data output from the image feature amount calculation device is the same as the resolution of the input image. For this reason, the memory of the feature data output device that receives the feature data can be reduced.
Further, after enlarging the input image, feature amount data is created, and feature regions are extracted based on the feature amount data. Thus, since the resolution of the feature amount data is excellent, the target feature region can be divided from other regions finely and accurately.
In particular, by calculating the feature quantities of edges and lines after enlarging the input image, it is possible to more accurately extract the positions of fine lines and slight flat areas between characters.
Further, the noise component can be removed by performing expansion and contraction of the feature region. Further, the feature amount data is reduced by the reduction processing means. For this reason, it is possible to reduce the memory of the device that receives the feature data.
The feature region is extracted based on the reduced feature value data. For this reason, the processing time concerning a characteristic area can be shortened. Further, by performing the reduction process based on the reciprocal of the enlargement magnification in the enlargement processing means, classification data having the same size as the input image can be obtained.

  The embodiment disclosed this time should be considered as illustrative in all points and not restrictive. The scope of the present invention is defined by the terms of the claims, rather than the description above, and is intended to include any modifications within the scope and meaning equivalent to the terms of the claims.

It is a block diagram which shows the hardware constitutions of the image characteristic area division | segmentation apparatus which concerns on 1st Embodiment. It is a block diagram which shows the hardware constitutions of the feature-value extraction process part of the image feature area division | segmentation apparatus which concerns on 1st Embodiment. It is a block diagram which shows the hardware constitutions of the area | region extraction process part of the image feature area division | segmentation apparatus which concerns on 1st Embodiment. It is a flowchart of the process of the image feature area division | segmentation apparatus which concerns on 1st Embodiment. It is a figure which shows the input image of 1st and 2nd embodiment. It is a figure which shows the calculation operator used with the image feature area division | segmentation apparatus of 1st and 2nd Embodiment and the image feature-value calculation apparatus of 3rd and 4th Embodiment. It is a figure which shows the feature-value data of 1st and 2nd embodiment. It is a figure which shows the classification data of 1st Embodiment. It is a figure which shows the area | region expansion data of 1st Embodiment. It is a figure which shows the classification data of 1st Embodiment. It is a figure which shows the area | region degeneration data of 1st Embodiment. It is a figure which shows the classification data of 1st Embodiment. It is a figure which shows the area | region division data of 1st Embodiment. It is a block diagram which shows the hardware constitutions of the image characteristic area division | segmentation apparatus which concerns on 2nd Embodiment. It is a block diagram which shows the hardware constitutions of the feature-value extraction process part of the image feature area division | segmentation apparatus which concerns on 2nd Embodiment. It is a block diagram which shows the hardware constitutions of the area | region extraction process part of the image feature area division | segmentation apparatus which concerns on 2nd Embodiment. It is a flowchart of the process of the image feature area division | segmentation apparatus which concerns on 2nd Embodiment. It is a figure which shows the classification data of 2nd Embodiment. It is a figure which shows the classification data of 2nd Embodiment. It is a figure which shows the area | region degeneration data of 2nd Embodiment. It is a figure which shows the classification data of 2nd Embodiment. It is a block diagram which shows the hardware constitutions of the image feature-value calculation processing apparatus which concerns on 3rd Embodiment. It is a flowchart of a process of the image feature-value calculation processing apparatus which concerns on 3rd Embodiment. It is a block diagram which shows the hardware constitutions of the image feature-value calculation processing apparatus which concerns on 4th Embodiment. It is a flowchart of a process of the image feature-value calculation processing apparatus which concerns on 4th Embodiment.

Explanation of symbols

  DESCRIPTION OF SYMBOLS 101 Image input device, 110,102 Image feature area dividing device, 103 Area division data output device, 104 Input image data buffer, 105,108 Feature quantity extraction process part, 106 Feature quantity data buffer, 107,109 Area extraction process part, 201 Enlargement processing unit, 202 Enlarged image data buffer, 203 Feature amount calculation processing unit, 301 Feature amount classification processing unit, 302 Classification data buffer, 303 Classification data decision determination processing unit, 304 Area deletion processing unit, 305 Data reduction processing unit, 306 Region expansion processing unit, 307 Region expansion data buffer, 308 Region reduction processing unit, 309 Region reduction data buffer, 604 Reduction processing unit, 902, 1102 Image feature amount calculation device, 903 Feature amount data output device

Claims (8)

An enlargement processing step for enlarging an input image including an object region in which a plurality of objects are arranged at predetermined intervals ;
In order to create feature amount data by calculating the feature amount of the image enlarged by the enlargement processing step for each pixel of the image using a plurality of arithmetic operators for the pixel and a predetermined range of pixels in the vicinity thereof. Feature amount calculation processing step,
Feature amount classification for creating classification data by classifying each pixel into a feature region and a non-feature region by thresholding each pixel of the feature amount data created in the feature amount calculation processing step Processing steps;
The feature area of the classification data created by the feature quantity classification processing step is integrated with at least a part of the feature area without integrating the objects of the object areas arranged at the predetermined interval. A first region expansion processing step for expanding the region with a first expansion amount for;
Among the small pieces constituting the feature area expanded in the first area expansion processing step, a small piece having a predetermined first size or more exceeding the size of the object arranged at the predetermined interval is extracted from the feature area. A first area deletion processing step for deletion;
A second expansion amount for integrating the objects in the object areas arranged at the predetermined intervals into the feature area in which the pieces of the first size or larger are deleted by the first area deletion processing step. A second region expansion processing step for expanding the region at
Of each of the small pieces constituting the feature region, a size equal to or smaller than a predetermined second size less than the size of the object region arranged at the predetermined interval integrated by the second region expansion processing step. An image feature region dividing program that causes a computer to execute a second region deletion processing step for deleting a small piece from the feature region. An enlargement processing step for enlarging an input image including an object region in which a plurality of objects are arranged at predetermined intervals ;
In order to create feature amount data by calculating the feature amount of the image enlarged by the enlargement processing step for each pixel of the image using a plurality of arithmetic operators for the pixel and a predetermined range of pixels in the vicinity thereof. Feature amount calculation processing step,
Feature amount classification for creating classification data by classifying each pixel into a feature region and a non-feature region by thresholding each pixel of the feature amount data created in the feature amount calculation processing step Processing steps;
The feature area of the classification data created by the feature quantity classification processing step is integrated with at least a part of the feature area without integrating the objects of the object areas arranged at the predetermined interval. A first region expansion processing step for expanding the region with a first expansion amount for;
Among the small pieces constituting the feature area expanded in the first area expansion processing step, a small piece having a predetermined first size or more exceeding the size of the object arranged at the predetermined interval is extracted from the feature area. A first area deletion processing step for deletion;
A second expansion amount for integrating the objects in the object areas arranged at the predetermined intervals into the feature area in which the pieces of the first size or larger are deleted by the first area deletion processing step. A second region expansion processing step for expanding the region at
Of each of the small pieces constituting the feature region, a size equal to or smaller than a predetermined second size less than the size of the object region arranged at the predetermined interval integrated by the second region expansion processing step. A second region deletion processing step for deleting a small piece from the feature region;
An image feature region dividing program causing a computer to execute a reduction processing step for reducing the size of the classification data. An enlargement processing step for enlarging an input image including an object region in which a plurality of objects are arranged at predetermined intervals;
In order to create feature amount data by calculating the feature amount of the image enlarged by the enlargement processing step for each pixel of the image using a plurality of arithmetic operators for the pixel and a predetermined range of pixels in the vicinity thereof. Feature amount calculation processing step,
A reduction processing step for reducing the size of the feature amount data created by the feature amount calculation processing step;
Feature amount classification processing step for classifying each pixel into a feature region and a non-feature region by thresholding each pixel of the feature amount data reduced in the reduction processing step and creating classification data When,
The feature area of the classification data created by the feature quantity classification processing step is integrated with at least a part of the feature area without integrating the objects of the object areas arranged at the predetermined interval. A first region expansion processing step for expanding the region with a first expansion amount for;
Among the small pieces constituting the feature area expanded in the first area expansion processing step, a small piece having a predetermined first size or more exceeding the size of the object arranged at the predetermined interval is extracted from the feature area. A first area deletion processing step for deletion;
A second expansion amount for integrating the objects in the object areas arranged at the predetermined intervals into the feature area in which the pieces of the first size or larger are deleted by the first area deletion processing step. A second region expansion processing step for expanding the region at
Of each of the small pieces constituting the feature region, a size equal to or smaller than a predetermined second size less than the size of the object region arranged at the predetermined interval integrated by the second region expansion processing step. An image feature region dividing program that causes a computer to execute a second region deletion processing step for deleting a small piece from the feature region. The image feature region dividing program according to claim 1, wherein the first expansion amount is 2 pixels. The said 1st area | region deletion process step counts the pixel number in the said small piece by 4 connection or 8 connection, and deletes the said small piece from the said characteristic area by this counted pixel number. The image feature region dividing program according to any one of the above. The image feature region dividing program according to claim 2 or 3, wherein in the reduction processing step, reduction processing is performed based on an inverse number of the enlargement magnification in the enlargement processing step. The method further includes a region degeneration processing step for degenerating the feature region after the region expansion processing in the second region expansion processing step and before the deletion processing in the second region deletion processing step. Item 7. The image feature region dividing program according to any one of Items 1 to 6. A computer-readable recording medium on which the image feature region dividing program according to claim 1 is recorded.

Images