JP4735168B2 - Identifying the subject expressed in the image - Google Patents

Description

  The present invention relates to an image processing technique, and more particularly to a technique for specifying a subject represented in an image.

  Attempts have been made to automatically identify what the subject represented in the image is. This technique is applied to, for example, an image processing apparatus that functions as an electronic pictorial book.

  Conventionally, a plurality of similarities are calculated using a feature amount extracted from a target image in which a subject is represented and a plurality of reference feature amounts corresponding to a plurality of objects prepared in advance, and high similarity is calculated. The object corresponding to the degree was selected as the subject candidate. When a plurality of objects (candidate objects) are selected as candidates for the subject, the user identifies what the subject is by confirming the detailed information of each object.

JP 2003-288363 A

  However, it has been difficult to identify the subject with the conventional technique. This is because it is necessary for the user to identify the subject by confirming the detailed information of each object, and it takes time and effort to confirm the detailed information of each object.

  The present invention has been made to solve the above-described problems in the prior art, and an object of the present invention is to provide a technique that can easily identify a subject represented in an image.

In order to solve at least a part of the problems described above, a first device of the present invention is an image processing device,
A plurality of reference feature data corresponding to a plurality of objects, each reference feature data indicating a feature of the corresponding object, and a reference feature data storage unit storing the plurality of reference feature data;
A difference data storage unit storing difference data related to differences between similar objects among the plurality of objects;
A feature data extraction unit that extracts feature data indicating features of the subject from a target image in which the subject is represented;
Using the feature data and the plurality of reference feature data corresponding to the plurality of objects, a candidate object as a candidate for the subject expressed in the target image is selected from the plurality of objects. A candidate selector to select;
When a plurality of candidate objects are selected, a difference information providing unit that provides a user with difference information indicated by the difference data between the plurality of candidate objects;
It is characterized by providing.

  In this apparatus, when a plurality of candidate objects are selected as subject candidates, difference information between the plurality of candidate objects is provided to the user, so that the user can easily select a subject based on the difference information. Can be identified.

In the first device, further
A plurality of detailed data corresponding to the plurality of objects, each detailed data indicating detailed information of the corresponding object, a detailed data storage unit storing the plurality of detailed data;
A detailed information providing unit that provides the user with the detailed information indicated by the detailed data corresponding to the designated candidate object when any of the plurality of candidate objects is designated by the user; ,
It is preferable to provide.

  In this way, the user can know the detailed information of the candidate object designated by himself / herself.

A second apparatus of the present invention is an image processing apparatus,
A plurality of reference feature data corresponding to a plurality of objects, each reference feature data indicating a feature of the corresponding object, and a reference feature data storage unit storing the plurality of reference feature data;
A difference data storage unit storing difference data related to differences between similar objects among the plurality of objects;
A subject identifying unit for identifying which of the plurality of objects is the subject represented in the target image;
With
The subject specifying unit
A feature data extraction unit that extracts feature data indicating the characteristics of the subject from the target image;
Using the feature data and the plurality of reference feature data corresponding to the plurality of objects, a candidate object as a candidate for the subject expressed in the target image is selected from the plurality of objects. A candidate selector to select;
A determination unit configured to determine any one of the plurality of candidate objects as the subject using the difference data between the plurality of candidate objects when the plurality of candidate objects are selected;
It is characterized by providing.

  In this apparatus, when a plurality of candidate objects are selected as subject candidates, difference data between the plurality of candidate objects is used by the determination unit, so that the subject can be easily identified and the subject Can be accurately identified.

In the second device:
The determination unit
According to the difference data between the plurality of candidate objects, the user is designated a reference image in which a specific part of the same or same kind of subject as the subject represented in the target image is represented. An instruction unit for instructing;
A reference feature data extraction unit that extracts reference feature data indicating characteristics of the specific part of the subject from the reference image designated by the user;
With
The determination unit
One of the plurality of candidate objects may be determined as the subject using the reference feature data and the plurality of reference feature data corresponding to the plurality of candidate objects.

  As described above, if the reference feature data is extracted from the reference image in which a specific part of the subject is expressed, the subject can be easily specified and the subject can be specified accurately.

In the second device,
A plurality of reference detailed feature data corresponding to the plurality of objects, wherein each of the reference detail feature data stores a plurality of reference detailed feature data indicating detailed features of a specific part of the corresponding object. Provided with a detailed reference feature data storage unit,
The determination unit
A reference detail feature data selection unit that selects a plurality of reference detail feature data corresponding to the plurality of candidate objects based on the difference data between the plurality of candidate objects;
A detailed feature data extraction unit that extracts detailed feature data indicating detailed features of the specific part of the subject from the target image;
With
The determination unit
Any of the plurality of candidate objects may be determined as the subject using the detailed feature data and the plurality of reference detailed feature data corresponding to the plurality of candidate objects.

  As described above, if a plurality of reference detailed feature data corresponding to a plurality of candidate objects are selected and the detailed feature data is extracted from the target image, the subject can be easily identified and the subject can be accurately identified. Can be specified.

In the second device:
The determination unit
Based on the difference data between the plurality of candidate objects, comprising a setting unit for setting a weight for determining any of the plurality of candidate objects as the subject,
The determination unit
According to the weighting setting, any one of the plurality of candidate objects is determined as the subject using the feature data and the plurality of reference feature data corresponding to the plurality of candidate objects. Also good.

  Thus, if weighting is set, the subject can be easily specified and the subject can be specified accurately.

In the above device,
The determination unit
Using the feature data and the plurality of reference feature data corresponding to the plurality of candidate objects, a plurality of similarities are calculated, and the candidate object corresponding to the maximum similarity among the plurality of similarities Determine an object as the subject,
The determination unit
A plurality of elements calculated using a plurality of element data corresponding to a plurality of feature elements included in the feature data and a plurality of reference element data corresponding to the plurality of feature elements included in the reference feature data Each of the similarities may be calculated by weighting the similarities.

In the second device,
A plurality of detailed data corresponding to the plurality of objects, each detailed data indicating detailed information of the corresponding object, a detailed data storage unit storing the plurality of detailed data;
Details that provide the user with the detailed information indicated by the detailed data corresponding to the determined object candidate when any of the plurality of candidate objects is determined as the subject by the determining unit An information provider,
It is preferable to provide.

  In this way, the user can know the detailed information of the subject determined by the determination unit.

  The present invention can be realized in various forms, for example, an image processing apparatus and method, a computer program for realizing the function of the method or apparatus, a recording medium storing the computer program, and the computer It can be realized in the form of a data signal including a program and embodied in a carrier wave.

A. First embodiment:
A-1. Configuration of image processing device:
Next, embodiments of the present invention will be described based on examples. FIG. 1 is an explanatory diagram showing an image processing apparatus (personal computer) 200 that functions as an electronic pictorial book. By using this device, when the user does not know what the subject (for example, a plant) represented in the target image is, the user can easily identify the subject and display detailed information about the subject. Obtainable.

  The image processing apparatus 200 includes a CPU 210, an internal storage device 220 such as a ROM and a RAM, an external storage device 250, a display unit 260, an operation unit 270 such as a mouse and a keyboard, and an interface unit (I / F unit) 290. And. The I / F unit 290 performs data communication with various devices provided outside. For example, the I / F unit 290 receives image data from the digital camera 400.

  The external storage device 250 stores a plurality of pictorial book databases 252 classified by categories such as plants, animals, and insects.

  FIG. 2 is an explanatory diagram showing the contents of one pictorial book database 252. FIG. 2 shows the contents of a plant pictorial book database.

  As shown in FIG. 2, one picture book database includes a plurality of object data corresponding to a plurality of objects. The object data is registered in association with the name of the object for each object (plant). The object data includes detailed data of the corresponding object and reference feature data of the corresponding object.

  The detailed data includes detailed information such as one or more images in which the corresponding object is expressed, and a detailed description about the corresponding object. The detailed description includes, for example, the form of the object (plant) (flower color / shape, leaf color / shape, stem color / shape, etc.), habitat, habitat environment, time of germination and flowering, etc. Includes a description.

  In the present embodiment, the detailed data further includes the names of other objects similar to the corresponding object. In FIG. 2, plants B and C are registered as plants similar to the plant A.

  The reference feature data is data indicating various features of the corresponding object. More specifically, the reference feature data includes a plurality of element data indicating a plurality of feature elements. The reference feature data includes, for example, a plurality of element data indicating a plurality of feature elements such as a flower color of an object (plant), a flower shape, a leaf color, and a leaf shape.

  In the present embodiment, values (frequency) of a plurality of classes in the color histogram are used as element data relating to color. As is well known, the color histogram is a histogram obtained by classifying each pixel in an image into a plurality of classes according to the color. For example, the element data relating to the color of the flower can be obtained by extracting a flower portion from a reference image in which a plant is expressed and obtaining a value of a color histogram using the extracted image. The same applies to the element data relating to the leaf color.

  In this embodiment, binary image data having a predetermined size (number of pixels) is used as the element data relating to the shape. For example, element data relating to the shape of a leaf is obtained by selecting a rectangular image surrounding a leaf portion from a reference image in which a plant is expressed, extracting a leaf contour by performing edge extraction processing on the rectangular image, It is obtained by reducing a binary image showing different values from the outside of the contour to a predetermined size. The same applies to the element data relating to the shape of the flower.

  The external storage device 250 further stores a plurality of difference databases 254 classified by categories such as plants, animals, and insects. Each difference database is used when a pictorial book database of the same category is used.

  FIG. 3 is an explanatory diagram showing the contents of one difference database 254. FIG. 3 shows the contents of the plant difference database corresponding to the plant pictorial book database.

  As shown in the drawing, the difference database includes difference data regarding differences between similar objects. The difference data includes a description of differences between similar objects (for example, plant A and plant B), an image of a first object (for example, plant A) relating to the difference, and a second object ( For example, difference information such as an image of plant B) is included. The description of the difference includes an explanation of points for distinguishing similar objects. In addition, the images of the respective objects related to the differences include images corresponding to explanations of points for distinguishing similar objects. For example, when the point that distinguishes two plants A and B is the shape of a leaf, the description of the difference includes the description of the shape of the leaf of both plants, and the image of each object related to the difference Includes an image representing the leaf shape of each plant.

  A computer program that functions as the image processing unit 300 is stored in the internal storage device 220 (FIG. 1). Note that the functions of the image processing unit 300 are realized by the CPU 210 executing a computer program. The computer program is provided in a form recorded on a computer-readable recording medium such as a CD-ROM.

  The image processing unit 300 includes a subject specifying unit 302 and a detailed information providing unit 304.

  The subject specifying unit 302 includes a feature data extracting unit 312, a candidate selecting unit 314, and a difference information providing unit 316. Which of the plurality of objects is the subject represented in the image? Is identified.

  The feature data extraction unit 312 analyzes target image data representing a target image in which a subject to be specified is expressed, and extracts feature data indicating the feature of the subject.

  The candidate selection unit 314 uses a plurality of feature data extracted from the target image data by the feature data extraction unit 312 and a plurality of reference feature data corresponding to a plurality of objects included in one picture book database 252. A candidate object as a candidate for the subject expressed in the target image is selected from the objects.

  The difference information providing unit 316 reads difference data between the plurality of candidate objects from the difference database 254 and displays the difference information on the display unit 260 when a plurality of candidate objects are selected by the candidate selecting unit 314. Let

  In this embodiment, when only one candidate object is selected by the candidate selection unit 314, the subject specifying unit 302 determines the one candidate object as a subject. When a plurality of candidate objects are selected by the candidate selection unit 314, the user determines one of the plurality of candidate objects as a subject.

  The detailed information providing unit 304 reads detailed data corresponding to the subject (object) determined by the subject specifying unit 302 or the user from the pictorial book database 252 and causes the display unit 260 to display the detailed information.

  In addition, the storage area in the external storage device 250 in which a plurality of reference feature data included in the picture book database 252 in the present embodiment is stored corresponds to the reference feature data storage unit in the present invention, and a plurality of information included in the picture book database 252 is included. The storage area in the external storage device 250 in which the detailed data is stored corresponds to the detailed data storage unit in the present invention. Further, the storage area in the external storage device 250 in which the difference database 254 in this embodiment is stored corresponds to the difference data storage unit in the present invention.

A-2. Processing of the image processing unit:
FIG. 4 is a flowchart showing a series of processing procedures of the image processing unit 300.

  In step S102, the user selects one pictorial book database from the plural pictorial book databases 252. The selection of the picture book database by the user is realized when the user selects a desired picture book from a plurality of picture books on the picture book selection screen prepared by the image processing unit 300. In addition, in a present Example, the case where the pictorial book of a plant is selected by the user is assumed.

  In step S104, the user designates a target image in which a subject to be identified is expressed. The designation of the target image by the user is realized, for example, when the user selects a desired image from a plurality of images on the target image selection screen prepared by the image processing unit 300. In this embodiment, plants to be specified are expressed in the target image.

  In step S106, the feature data extraction unit 312 analyzes the target image data representing the target image, and extracts feature data indicating the characteristics of the subject. The extracted feature data corresponds to the reference feature data of the pictorial book database used. In this embodiment, the plant picture book database 252 is selected, and the reference feature data of the database includes, as shown in FIG. 2, information such as flower color, flower shape, leaf color, leaf shape, etc. A plurality of element data indicating a plurality of feature elements are included. Therefore, in this embodiment, feature data including a plurality of element data indicating a plurality of feature elements such as a flower color, a flower shape, a leaf color, and a leaf shape is extracted from the target image data.

  Specifically, the feature data extraction unit 312 first determines an area where each part of the subject (plant) is expressed in the target image. For example, an edge image is generated by performing an edge extraction process on the target image, and the target image is divided into a plurality of regions based on the edge image. Of the plurality of regions, a region having a high possibility of being a flower color such as red, white, and yellow is determined as a flower region. In addition, an area that is highly likely to be a leaf color such as green is determined as a leaf area. If it is difficult to determine the area, the user may be instructed to specify a flower area or a leaf area.

  Next, the feature data extraction unit 312 generates element data. For example, the element data related to the color of the flower is generated by classifying each pixel constituting the flower region into a plurality of classes according to the color and obtaining the value (frequency) of the color histogram. The same applies to the element data relating to the leaf color. In addition, the element data relating to the shape of the flower is selected by selecting a rectangular image surrounding the flower region, and by performing edge extraction processing on the rectangular image, the contour of the flower is extracted and shows different values between the inside of the contour and the outside of the contour. It is generated by reducing the binary image to a predetermined size. The same applies to the element data relating to the shape of the leaves.

  In addition, when only a part of the subject (for example, a flower of a plant) is expressed in the target image, the feature data includes only a part of element data (element data such as flower color and flower shape). And other element data (element data such as leaf color and leaf shape) are not included.

  In step S108, the candidate selection unit 314 selects a candidate for the subject from among a plurality of objects registered in the pictorial book database selected in step S102.

  Specifically, the candidate selection unit 314 first obtains the feature data obtained in step S106 and a plurality of reference feature data corresponding to a plurality of objects registered in the pictorial book database selected in step S102. To calculate a plurality of similarities.

  FIG. 5 is an explanatory diagram illustrating an example of a similarity calculation result. As described above, in the present embodiment, the feature data and the reference feature data each include a plurality of feature elements indicating a plurality of feature elements of a flower color, a flower shape, a leaf color, and a leaf shape. Contains element data. Thus, when the feature data and the reference feature data include a plurality of element data, the element similarity is calculated for each corresponding feature element, and the similarity is obtained by obtaining an average value of the plurality of element similarities. Calculated.

  In addition, when only a part of the subject (for example, a flower of a plant) is expressed in the target image, a part of element data (element data such as flower color and flower shape) included in the feature data is included. Only the corresponding partial element similarity is calculated, and the similarity is calculated by obtaining an average value of the partial element similarity.

  In this embodiment, the element similarity regarding the color is calculated using the angular distance cos θ. The angular distance cos θ is expressed by the following equation (1).

Here, the vector I and the vector O are vectors indicating two element data that are calculation targets of the element similarity regarding the color. If the values of the classes h1, h2, h3,... Of the color histogram are vh1, vh2, vh3,..., Each vector is represented by (vh1, vh2, vh3,...), And the magnitude of each vector is (vh1 ² + Vh2 ² + vh3 ² +...) ^1/2 .

  For example, when calculating the element similarity regarding the color of a flower, a vector indicating element data of the subject flower color and a vector indicating element data of the object flower color are used as the vector I and the vector O, respectively. And are used.

  In the present embodiment, the element similarity related to the shape is calculated using two binary image data that are two element data to be calculated. Specifically, two binary images are overlaid, and at least one region of the first region inside the contour of the first binary image and the second region inside the contour of the second binary image is determined. The area of the included region (first pixel number) and the area of the overlapping portion of the first region and the second region (second pixel number) are calculated. Then, the value of (second pixel number) / (first pixel number) is calculated as the element similarity regarding the shape. When the element similarity regarding the shape is calculated, one of the two binary images is preferably translated or rotated so that the element similarity is maximized.

  For example, when calculating the element similarity regarding the leaf shape, element data (binary image data) of the shape of the subject leaf is used as the first binary image data, and the second binary image data is used as the second binary image data. Element data (binary image data) of the leaf shape of the object is used.

  When a plurality of similarities corresponding to a plurality of objects are calculated in step S108 (FIG. 4), the candidate selecting unit 314 next selects a subject candidate using the plurality of similarities. If there is another similarity that is approximately equal to the maximum similarity among the plurality of similarities (for example, the difference between the maximum similarity and the other similarity is 0.1 or less), the maximum A plurality of objects corresponding to the similarity and other similarities are selected as subject candidates. If there are no other similarities in the multiple similarities that are approximately equal to the maximum similarity, in other words, if the other similarities are much smaller than the maximum similarity (eg, the maximum similarity) When the difference between the degree and another similarity exceeds 0.1), one object corresponding to the maximum degree of similarity is selected as a subject candidate. However, when all the similarities are a relatively small predetermined value (for example, 0.5) or less, the candidate selection unit 314 does not select a subject candidate.

  In step S110, the candidate selection unit 314 determines whether a subject candidate has been selected in step S108. If a subject candidate is selected, the process proceeds to step S112. On the other hand, when the subject candidate is not selected, the processing in FIG. 4 ends. In this case, it is preferable that the subject specifying unit 302 notifies the user that no subject candidate has been selected.

  In step S112, the candidate selection unit 314 determines whether only one candidate object is selected as a subject candidate in step S108. When only one candidate object is selected as the subject candidate, the process proceeds to step S114, and when a plurality of candidate objects are selected as the subject candidates, the process proceeds to step S212.

  In step S114, the subject specifying unit 302 determines the only candidate object selected as the subject candidate as the subject.

  In step S116, the detailed information providing unit 304 reads the detailed data of the candidate object determined as the subject by the subject specifying unit 302 from the pictorial book database 252 selected in step S102, and the details represented by the detailed data. Information is displayed on the display unit 260.

  In step S212, the difference information providing unit 316 acquires difference data between a plurality of candidate objects selected as subject candidates from the difference database 254 in the same category as the pictorial book database 252 selected in step S102. Then, the difference information providing unit 316 displays the difference information represented by the difference data on the display unit 260.

  FIG. 6 is an explanatory diagram showing an example of the difference information display screen W displayed in step S212 (FIG. 4). FIG. 6 shows a screen displayed when plant A and plant B are selected as subject candidates.

  As shown in the figure, the difference information display screen W includes a first field F1 for displaying an explanation of the difference between the plant A and the plant B, and a second field F2a for displaying an image of the plant A regarding the difference. A third field F2b for displaying an image of the plant B related to the difference is provided.

  From the description of the difference displayed in the first field F1, the user can know that the point that distinguishes the plant A from the plant B is the back of the leaf. Further, from the images of the plant A and the plant B related to the differences displayed in the second and third fields F2a and F2b, the user determines the mode of the back side of the leaf of the plant A and the mode of the back side of the leaf of the plant B. I can know.

  In step S214 (FIG. 4), the user designates one of the plurality of candidate objects selected as the subject candidate as the subject. The difference information display screen W of FIG. 6 is provided with a button B1a for displaying detailed information of the plant A and a button B2a for displaying detailed information of the plant B. Based on the difference information displayed in the three fields F1, F2a, and F2b of the difference information display screen W, the user selects any candidate object that is considered to be a subject and sets the two candidate objects. One of the two corresponding buttons B1a and B1b is selected.

  In step S116, the detailed information providing unit 304 reads the detailed data of the candidate object determined as the subject by the user from the pictorial book database 252 selected in step S102, and displays the detailed information represented by the detailed data. Displayed on the unit 260. Thereby, the user can know the detailed information of the subject.

  By the way, in this embodiment, as shown in FIG. 6, the difference information display screen W displays an object (plant) similar to the candidate object (plant). As described in FIG. 2, the plant B and the plant C are registered in the detailed data of the plant A as plants similar to the plant A. Moreover, the case where the plant A and the plant C are registered into the detailed data of the plant B as plants similar to the plant B is assumed. That is, the case where the plant C is similar to the plant A and the plant C is assumed. As described above, in this embodiment, plant A and plant B are selected as subject candidates. For this reason, on the difference information display screen W, a plant C similar to both the plant A and the plant B selected as the subject candidates is displayed as a similar plant.

  In addition, on the difference information display screen W of FIG. 6, three buttons B2a to B2c are prepared for the user to obtain information on an object similar to the candidate object. When the button B2a is selected, the detailed information providing unit 304 reads the detailed data of the plant C from the plant pictorial book database and causes the display unit 260 to display the detailed information. When the button B2b is selected, the difference information providing unit 316 reads the difference data between the plant A and the plant C from the plant difference database and causes the display unit 260 to display the difference information. Similarly, when the button B2c is selected, the difference information providing unit 316 reads the difference data between the plant B and the plant C from the plant difference database and causes the display unit 260 to display the difference information.

  In the present embodiment, the case where two candidate objects are selected as subject candidates has been described. However, three or more candidate objects may be selected. In this case, for example, the difference information for each of the two candidate objects may be provided to the user. For example, when three candidate objects A to C are selected, difference information between candidate objects A and B, difference information between candidate objects B and C, and between candidate objects A and C The difference information may be provided to the user.

  As described above, in the present embodiment, when a plurality of candidate objects are selected as subject candidates, difference information between the plurality of candidate objects is provided to the user. The subject can be easily specified based on the above.

B. Second embodiment:
FIG. 7 is an explanatory diagram showing an image processing apparatus (personal computer) 200B in the second embodiment. FIG. 7 is substantially the same as FIG. 1, but the difference database 254B and the image processing unit 300B are changed.

  FIG. 8 is an explanatory diagram showing the contents of one difference database 254B. FIG. 8 shows the contents of the plant difference database corresponding to the plant pictorial book database, as in FIG. 3. As can be seen from a comparison between FIG. 8 and FIG. 3, also in the present embodiment, the difference database includes difference data regarding differences between similar objects. However, in the present embodiment, unlike the first embodiment, the difference data includes differences between similar objects (for example, the plant A and the plant B). Does not include images of each object. Differences may be represented by, for example, “flowers” or “leaves”, or “flower colors”, “flower shapes”, “leaf colors”, or “leaf shapes”. May be.

  In the present embodiment, the image processing unit 300B includes a subject specifying unit 302B including a determining unit 320. The determination unit 320 includes a reference image designation instruction unit 322 and a reference feature data extraction unit 324. When a plurality of candidate objects are selected by the candidate selection unit 314, the plurality of candidate objects are selected. One of them is determined as a subject.

  When the candidate selection unit 314 selects a plurality of candidate objects, the reference image designation instruction unit 322 reads difference data between the plurality of candidate objects from the difference database 254B. Then, the reference image designation instruction unit 322 instructs the user to designate a reference image in which a specific part of the subject that is the same as or similar to the subject represented in the target image is represented according to the difference data. Is displayed on the display unit 260. The reference image is an image in which the difference indicated by the difference data is expressed in detail.

  The reference feature data extraction unit 324 analyzes reference image data representing a reference image designated by the user according to the difference data, and extracts reference feature data indicating the characteristics of a specific part of the subject. The reference feature data includes a plurality of element data related to the difference indicated by the difference data.

  FIG. 9 is a flowchart showing a series of processing procedures of the image processing unit 300B. FIG. 9 is substantially the same as FIG. 4, but steps S222 to S228 are executed instead of steps S212 and S214.

  In step S222, the reference image designation instruction unit 322 acquires difference data between a plurality of candidate objects selected as subject candidates from the difference database 254B of the same category as the picture book database 252 selected in step S102. Then, the reference image designation instruction unit 322 causes the display unit 260 to display a message instructing the user to designate a reference image in which a difference (for example, “flower shape”) is expressed according to the difference data.

  In step S224, the user designates a reference image in which a specific part of the subject identical or similar to the subject represented in the target image is expressed in detail. The designation of the reference image by the user is realized, for example, when the user selects an appropriate image from a plurality of images on the reference image selection screen prepared by the reference image designation instruction unit 322. For example, when instructed to specify a reference image in which “flower shape” is expressed in step S222, the user selects “flower shape” of a subject that is the same as or similar to the subject expressed in the target image. "Is specified.

  In step S226, the reference feature data extraction unit 324 analyzes reference image data representing a reference image and extracts reference feature data indicating the characteristics of a specific part of the subject, in the same manner as the feature data extraction in step S106. . The extraction of the reference feature data is performed based on the difference data acquired in step S222. For example, when the difference indicated by the difference data is “flower shape”, the reference feature data extraction unit 324 extracts reference feature data including element data relating to the shape of the flower. When the difference is “flower”, the reference feature data extraction unit 324 extracts reference feature data including element data relating to the color of the flower and element data relating to the shape of the flower.

  In step S228, the determination unit 320 determines one of the plurality of candidate objects selected in step S108 as a subject. Specifically, the determination unit 320 calculates a plurality of reference similarities using the reference feature data obtained in step S226 and a plurality of reference feature data corresponding to a plurality of candidate objects. The reference similarity is calculated in the same manner as the similarity calculation in step S108. However, when the difference is the “flower shape”, only element data relating to the flower shape of the reference feature data corresponding to each candidate object is used. When the difference is “flower”, only the element data related to the color of the flower and the element data related to the shape of the flower among the reference feature data corresponding to each candidate object are used. Then, the determination unit 320 determines one candidate object corresponding to the maximum reference similarity among the plurality of candidate objects as a subject.

  When three or more candidate objects are selected as subject candidates, for example, a reference image may be designated for each of two candidate objects. For example, when three candidate objects A to C are selected, the first reference image is designated based on the difference data between the candidate objects A and B. Then, two reference similarities corresponding to the two candidate objects A and B are calculated, and one candidate object (for example, candidate object B) remains as a subject candidate. Next, a second reference image is designated based on the difference data between candidate objects B and C. Then, two reference similarities corresponding to the two candidate objects B and C are calculated, and one candidate object is determined as a subject. Note that when the difference data of the candidate objects A to C indicate the same difference (for example, “flower shape”), one reference image is designated. Then, three reference similarities corresponding to the three candidate objects A to C are calculated, and any one candidate object is determined as a subject.

  As described above, in this embodiment, when a plurality of candidate objects are selected as subject candidates, reference feature data is extracted from the reference image according to the difference data. Then, using the reference feature data and the plurality of reference feature data corresponding to the plurality of candidate objects, one candidate object is determined as a subject. For this reason, the subject can be easily specified, and the subject can be accurately specified.

  In the present embodiment, for example, only a part of the subject (for example, a flower of a plant) is represented in the target image, and the difference between the plurality of candidate objects is the other part (for example, the shape of a plant leaf). Is also available. Further, in the present embodiment, for example, one side of a subject (for example, a side surface of a plant) is expressed in the target image, and a portion where a difference between a plurality of candidate target objects is not expressed in the target image (for example, a flower It can also be used in the case of a front shape).

C. Third embodiment:
FIG. 10 is an explanatory diagram showing an image processing apparatus (personal computer) 200C in the third embodiment. FIG. 10 is substantially the same as FIG. 1, but a plurality of detailed pictorial book databases 256 are added to the external storage device 250.

  As described above, each picture book database 252 is divided into large categories such as plants, animals, and insects. On the other hand, each detailed picture book database 256 is divided into small categories (parts) such as plant flowers and plant leaves.

  FIG. 11 is an explanatory diagram showing the contents of one detailed pictorial book database 256. In FIG. 11, the contents of the detailed flower book database 256 of plant flowers are shown.

  As shown in FIG. 11, one detailed pictorial book database includes a plurality of object part data corresponding to a plurality of objects. The object part data is registered in association with the name of the object for each object (plant). The object part data includes detailed data of a specific part of the corresponding object and reference feature data of the specific part of the corresponding object (hereinafter referred to as “reference detailed feature data”).

  The detailed data includes detailed information such as one or more images in which a specific part of the corresponding target object is expressed, and a detailed description regarding the specific part of the corresponding target object. The detailed description includes, for example, a description of an aspect of a specific part (flower) of the object (plant).

  The detailed data in the detailed picture book database may include the same data as the detailed data in the picture book database, or may include different data.

  The reference detailed feature data includes a plurality of element data indicating a plurality of feature elements of a specific part of the corresponding object. The reference detailed feature data includes, for example, a plurality of element data indicating features such as the color of the object (plant), the shape of the flower, the shape of the petals, and the number of petals.

  In the present embodiment, binary image data having a predetermined size (number of pixels) is used as the element data related to the shape of the petals, similarly to the element data related to the other shapes described above. Element data related to the shape of the petal is selected from a reference image representing the flower of the plant, a rectangular image surrounding the petal portion is selected, and an edge extraction process is performed on the rectangular image to extract the outline of the petal. It is obtained by reducing a binary image showing different values from the outside of the contour to a predetermined size.

  Further, in the present embodiment, previously investigated values are used as element data relating to the number of petals.

  The reference detailed feature data in the detailed picture book database may include the same data as the reference feature data in the picture book database as element data related to the color of the flower or the shape of the flower, or may include different data. Also good.

  Further, in this embodiment, the difference database 254B and the image processing unit 300C are changed so as to be understood by comparing FIG. 10 and FIG. The difference database 254B of this embodiment is the same as that of the second embodiment (FIG. 8).

  In this embodiment, the image processing unit 300C includes a subject specifying unit 302C including a determining unit 330. The determination unit 330 includes a difference data acquisition unit 332, a detailed picture book database selection unit 334, and a detailed feature data extraction unit 336. When a plurality of candidate objects are selected by the candidate selection unit 314, One of the plurality of candidate objects is determined as a subject.

  When the candidate selection unit 314 selects a plurality of candidate objects, the difference data acquisition unit 332 reads the difference data between the plurality of candidate objects from the difference database 254B.

  The detailed picture book database selection unit 334 selects one detailed picture book database from a plurality of detailed picture book databases 256 based on the difference data. For example, when the difference indicated by the difference data is “flower”, “flower color”, or “flower shape”, the detailed picture book database selection unit 334 selects the detailed picture book database of plant flowers. . Further, the detailed picture book database selection unit 334 reads a plurality of reference detailed feature data corresponding to the plurality of candidate objects selected by the candidate selection part 314 from the selected detailed picture book database.

  Similar to the feature data extraction unit 312, the detailed feature data extraction unit 336 analyzes the target image data representing the target image in which the subject is represented, and extracts detailed feature data indicating the detailed features of a specific part of the subject. To do.

  The storage area in the external storage device 250 in which the reference detailed feature data included in the detailed pictorial book database 256 in this embodiment is stored corresponds to the reference detailed feature data storage unit in the present invention. Further, the difference data acquisition unit 332 and the detailed pictorial book database selection unit 334 in the present embodiment correspond to the reference detailed feature data selection unit in the present invention.

  FIG. 12 is a flowchart showing a series of processing procedures of the image processing unit 300C. FIG. 12 is substantially the same as FIG. 4, but steps S232 to S238 are executed instead of steps S212 and S214.

  In step S232, the difference data acquisition unit 332 acquires difference data between a plurality of candidate objects selected as subject candidates from the difference database 254B of the same category as the pictorial book database 252 selected in step S102.

  In step S234, the detailed picture book database selection unit 334 selects a detailed picture book database based on the difference data acquired in step S232. In this embodiment, it is assumed that the difference indicated by the difference data is “flower”, and the detailed picture book database selection unit 334 selects the detailed picture book database 256 of the plant flower. Further, the detailed pictorial book database selection unit 334 reads a plurality of reference detailed feature data corresponding to the plurality of candidate objects selected in step S108 from the selected detailed pictorial book database.

  In step S236, as in step S106, the detailed feature data extraction unit 336 analyzes the target image data representing the target image and extracts the detailed feature data from the target image data. The extracted detailed feature data corresponds to the reference detailed feature data of the detailed pictorial book database to be used. In this embodiment, the plant detailed flower book database 256 is selected, and the reference detailed feature data of the database includes, as shown in FIG. 11, the color of the flower, the shape of the flower, the shape of the petal, and the petal. A plurality of element data indicating a plurality of characteristic elements, such as the number of sheets, is included. Therefore, in this embodiment, detailed feature data including a plurality of element data indicating a plurality of feature elements such as a flower color, a flower shape, a petal shape, and the number of petals is extracted from the target image data. .

  Specifically, like the feature data extraction unit 312, the detailed feature data extraction unit 336 first determines an area in which each part of the subject (plant) is expressed in the target image. The petal region can be determined as follows, for example. An edge image is generated by performing an edge extraction process on the target image, and the target image is divided into a plurality of regions based on the edge image. Of the plurality of regions, two or more regions that are highly likely to have petal colors such as red, white, and yellow and that have substantially the same area (number of pixels) are determined as petal regions. If it is difficult to determine the petal region, the user may indicate the petal region. Next, the detailed feature data extraction unit 336 generates element data in the same manner as the feature data extraction unit 312. The element data relating to the number of petals can be generated, for example, by counting the number of areas determined as petal areas.

  In step S238, the determination unit 330 determines one of the plurality of candidate objects selected in step S108 as a subject. Specifically, the determination unit 330 calculates a plurality of detailed similarities using the detailed feature data obtained in step S236 and the plurality of reference detailed feature data corresponding to the plurality of candidate objects. The detailed similarity is calculated in the same manner as the similarity calculation in step S108. The element similarity regarding the number of petals is calculated by “(x− | xy −) / x” where x is the number of petals of the subject and y is the number of petals of the object candidate. Then, the determination unit 330 determines one candidate object corresponding to the maximum detailed similarity among the plurality of candidate objects as a subject.

  In step S116, the detailed information providing unit 304 reads detailed data of the candidate object determined as the subject by the determining unit 330 from the pictorial book database 252 selected in step S102, and is represented by the detailed data. Detailed information is displayed on the display unit 260. Thereby, the user can know the detailed information of the subject. In this embodiment, a detailed picture book database 256 is used. For this reason, the detailed information providing unit 304 reads out detailed data regarding each part of the candidate object determined as the subject from the detailed pictorial book database 256 together with the above detailed information, and displays the detailed information represented by the detailed data You may display on the part 260. FIG.

  When three or more candidate objects are selected as subject candidates, for example, two reference detailed feature data are selected for each of the two candidate objects, and detailed feature data is extracted from the target image. It only has to be done. For example, when three candidate objects A to C are selected, two reference detailed feature data corresponding to the two candidate objects A and B are obtained based on the difference data between the candidate objects A and B. While being selected, the first detailed feature data is extracted from the target image. Then, two detailed similarities corresponding to the two candidate objects A and B are calculated, and one candidate object (for example, candidate object B) remains as a subject candidate. Next, based on the difference data between the candidate objects B and C, two reference detailed feature data corresponding to the two candidate objects B and C are selected, and the second detailed feature data is selected from the target image. Extracted. Then, two detailed similarities corresponding to the two candidate objects B and C are calculated, and one candidate object is determined as a subject. When the difference data of the candidate objects A to C indicate the same difference (for example, “flower shape”), three reference detailed feature data corresponding to the three candidate objects A to C are selected. At the same time, one detailed feature data is extracted from the target image. Then, three detailed similarities corresponding to the three candidate objects A to C are calculated, and any one candidate object is determined as a subject.

  As described above, in this embodiment, when a plurality of candidate objects are selected as subject candidates, a plurality of reference detailed feature data corresponding to the plurality of candidate objects are selected based on the difference data. In addition, detailed feature data is extracted from the target image. Then, using the detailed feature data and the plurality of reference detailed feature data, one candidate object is determined as a subject. For this reason, the subject can be easily specified, and the subject can be accurately specified.

  In this embodiment, the plant pictorial book database includes a plurality of reference feature data, and the plant flower detailed pictorial book database includes a plurality of reference detailed feature data. One picture book database may include both a plurality of reference feature data and a plurality of reference detailed feature data. In this case as well, a subject candidate is selected using the feature data and the plurality of reference feature data, and one of the plurality of candidate objects is selected using the detailed feature data and the plurality of reference detail feature data. What is necessary is just to be determined as a to-be-photographed object.

D. Fourth embodiment:
FIG. 13 is an explanatory diagram showing an image processing apparatus (personal computer) 200D in the fourth embodiment. FIG. 13 is almost the same as FIG. 1, but the difference database 254B and the image processing unit 300B are changed. The difference database 254B of this embodiment is the same as that of the second embodiment (FIG. 8).

  In the present embodiment, the image processing unit 300D includes a subject specifying unit 302D including a determining unit 340. The determination unit 340 includes a difference data acquisition unit 342 and a weighting setting unit 344. When a plurality of candidate objects are selected by the candidate selection unit 314, one of the plurality of candidate objects is selected. Is determined as the subject.

  Similar to the difference data acquisition unit 332 of the third embodiment, the difference data acquisition unit 342 selects a plurality of candidate objects from the difference database 254B when the candidate selection unit 314 selects a plurality of candidate objects. Read the difference data.

  More specifically, the weighting setting unit 344 determines, based on the difference data, which of the plurality of candidate objects is the subject represented in the target image. A weight is set for calculating a plurality of similarities corresponding to.

  Note that the difference data acquisition unit 342 and the weighting setting unit 344 in the present embodiment correspond to the setting unit in the present invention.

  FIG. 14 is a flowchart showing a series of processing procedures of the image processing unit 300D. FIG. 14 is substantially the same as FIG. 4, but steps S242 to S246 are executed instead of steps S212 and S214.

  In step S242, the difference data acquisition unit 342 acquires difference data between a plurality of candidate objects selected as subject candidates from the difference database 254B of the same category as the pictorial book database 252 selected in step S102.

  In step S244, the weight setting unit 344 sets a weight (weighting coefficient) for each element similarity used when calculating the similarity based on the difference data. For example, when the difference between two candidate objects selected as subject candidates is a specific part, the weight (weight coefficient) of the element similarity regarding the specific part is set to be relatively high. The element similarity weighting (weighting coefficient) is set relatively low.

  In step S246, the determination unit 340 calculates a plurality of reference similarities corresponding to a plurality of candidate objects based on the weight set in step S244. Specifically, the determination unit 340 includes the plurality of element similarities used when calculating the similarity in step S108 and the weight (weighting coefficient) set for each element similarity in step S244. The reference similarity is calculated for each candidate object. Then, the determination unit 340 determines the candidate object corresponding to the maximum reference similarity among the plurality of reference similarities corresponding to the plurality of candidate objects as the subject.

  FIG. 15 is an explanatory diagram showing the reference similarity calculated in step S246 (FIG. 14). FIG. 15A shows a target image in which a subject (plant) is expressed. FIGS. 15B-1 and 15B-2 show an object (plant) A and an object (plant) B selected as subject candidates, respectively. In this embodiment, as can be seen from FIGS. 15B-1 and 15B-2, the difference between the two objects is the “flower shape”.

  FIGS. 15C-1 and 15C-1 show the similarity calculated in step S108. The similarity shown in FIG. 15C-1 is calculated using the feature data of the target image and the reference feature data of the target A, and the similarity shown in FIG. 15C-2 is the feature of the target image. It is calculated using the data and the reference feature data of the object B. As described above, each similarity is calculated by averaging a plurality of element similarities. As can be seen by comparing FIGS. 15C-1 and 15C-2, the two similarities have almost the same value.

  FIGS. 15D-1 and 15D-2 show the reference similarity calculated in step S246. The reference similarity shown in FIG. 15 (D-1) is calculated using the feature data of the target image and the reference feature data of the target A as in FIG. 15 (C-1). The reference similarity shown in 2) is calculated using the feature data of the target image and the reference feature data of the target B as in FIG. However, the two reference similarities shown in FIGS. 15D-1 and 15D-2 are calculated by obtaining an average value of a plurality of element similarities multiplied by a weighting coefficient. Specifically, since the difference indicated by the difference data is “flower shape”, the weight coefficient “1.5” of the element similarity related to the flower shape is the weight coefficient “1. It is set larger than “0”. For this reason, as shown in FIGS. 15D-1 and 15D-2, the difference between the two reference similarities is considerably large.

  In this way, in step S246, two reference similarities corresponding to the two candidate objects are calculated, and the determination unit 340 selects the object A corresponding to the large reference similarity of the two reference similarities. Determine as the subject.

  When three or more candidate objects are selected as subject candidates, for example, weighting may be set for each of two candidate objects. For example, when three candidate objects A to C are selected, the first weight is set based on the difference data between the candidate objects A and B. Then, two reference similarities corresponding to the two candidate objects A and B are calculated, and one candidate object (for example, candidate object B) remains as a subject candidate. Next, the second weighting is designated based on the difference data between the candidate objects B and C. Then, two reference similarities corresponding to the two candidate objects B and C are calculated, and one candidate object is determined as a subject. When the difference data of the candidate objects A to C indicate the same difference (for example, “flower shape”), one type of weighting may be set. Then, three reference similarities corresponding to the three candidate objects A to C are calculated, and any one candidate object is determined as a subject.

  As described above, in the present embodiment, when a plurality of candidate objects are selected as subject candidates, weighting is set according to the difference data. Then, according to the weighting setting, one object candidate is determined as a subject using the feature data and a plurality of reference feature data corresponding to the plurality of candidate objects. For this reason, the subject can be easily specified, and the subject can be accurately specified.

  In addition, this invention is not restricted to said Example and embodiment, In the range which does not deviate from the summary, it can be implemented in a various aspect, For example, the following deformation | transformation is also possible.

(1) In the above-described embodiment, the color-related similarity is calculated using the angular distance (formula (1)) of the two vectors I and O indicating the two element data, but is calculated using another method. May be. For example, it may be calculated using the Euclidean distance between the two element data, more specifically, between the end points of the two vectors I and O.

  In the above embodiment, binary image data is used as element data relating to the shape. However, the element data relating to the shape may be expressed by a numerical value. In this case, the degree of similarity can be calculated using the angular distance (formula (1)) in the same manner as the degree of similarity regarding the color.

(2) In the second to fourth embodiments, as shown in FIG. 8, the difference data includes only the differences. Further, as in the first embodiment (FIG. 3), the differences are explained. And an image of each object related to the difference may be included. Similar to the first embodiment, difference information including a description of the difference may be provided to the user. In this aspect, the process of determining any of the plurality of candidate objects as the subject may be executed by the user or by the image processing unit (determination unit) according to the user's selection. Good.

  In the third example (FIG. 12), detailed feature data is extracted from the target image in step S236. However, as in the second example, the same or the same type as the subject expressed in the target image. Detailed feature data may be extracted from a reference image in which a specific part of the subject is expressed in detail.

  In general, when a plurality of candidate objects are selected as subject candidates, if any one of the plurality of candidate objects is determined as a subject using difference data between the plurality of candidate objects. Good.

(3) In the above embodiment, in step S108, candidate candidates are selected using feature data and a plurality of reference feature data corresponding to a plurality of objects, but prior to the processing in step S108. It is preferable that processing for reducing the number of objects to be investigated is executed.

  For example, the flowering time of plant flowers is often limited. For this reason, when time information indicating the time (date and time) at the time of shooting is added to the target image data, the target information to be investigated is obtained using the time information and information on the flowering time included in the detailed data. The number of objects can be quickly reduced. Specifically, when a plant flower is represented in the target image, if the time information of the target image data matches the information of the flowering time included in the detailed data of each target, step S108. If the object is not matched, it is excluded from the investigation target in step S108. In addition, the habitat of plants is often limited. For this reason, when position information (GPS information) indicating the position at the time of shooting is added to the target image data, the position information and the habitat information included in the detailed data are used to investigate The number of objects can be quickly reduced. Specifically, when the plant is represented in the target image, if the position information of the target image data matches the habitat information included in the detailed data of each target, the investigation in step S108 is performed. If it remains as a target object and does not match, it is excluded from the investigation target in step S108.

(4) In the above embodiment, a part of the configuration realized by hardware may be replaced by software, and conversely, a part of the configuration realized by software may be replaced by hardware.

It is explanatory drawing which shows the image processing apparatus 200 which functions as an electronic pictorial book. It is explanatory drawing which shows the content of the one pictorial book database 252. FIG. It is explanatory drawing which shows the content of one difference database. 5 is a flowchart showing a series of processing procedures of the image processing unit 300. It is explanatory drawing which shows the example of the calculation result of similarity. It is explanatory drawing which shows an example of the difference information display screen W displayed by step S212 (FIG. 4). It is explanatory drawing which shows the image processing apparatus 200B in 2nd Example. It is explanatory drawing which shows the content of one difference database 254B. It is a flowchart which shows a series of processing procedures of the image process part 300B. It is explanatory drawing which shows 200 C of image processing apparatuses in 3rd Example. It is explanatory drawing which shows the content of the one detailed illustration book database 256. FIG. It is a flowchart which shows a series of processing procedures of the image processing part 300C. It is explanatory drawing which shows image processing apparatus 200D in 4th Example. It is a flowchart which shows a series of processing procedures of image processing part 300D. It is explanatory drawing which shows the reference similarity calculated by step S246 (FIG. 14).

Explanation of symbols

200, B, C, D ... Image processing device (personal computer)
210 ... CPU
DESCRIPTION OF SYMBOLS 220 ... Internal storage device 250 ... External storage device 252 ... Picture book database 254, B ... Difference database 256 ... Detailed picture book database 260 ... Display part 270 ... Operation part 290 ... I / F part 300, B, C, D ... Image processing part 302, B, C, D ... Subject specifying unit 304 ... Detailed information providing unit 312 ... Feature data extracting unit 314 ... Candidate selection unit 316 ... Difference information providing unit 320 ... Determining unit 322 ... Reference image designation instruction unit 324 ... Reference feature data Extraction unit 330 ... determination unit 332 ... difference data acquisition unit 334 ... detailed picture book database selection unit 336 ... detailed feature data extraction unit 340 ... determination unit 342 ... difference data acquisition unit 344 ... weighting setting unit 400 ... digital camera

Claims (8)

An image processing apparatus,
A plurality of reference feature data corresponding to a plurality of objects, each reference feature data indicating a feature of the corresponding object, and a reference feature data storage unit storing the plurality of reference feature data;
Of the plurality of objects corresponding to the plurality of reference characteristic data stored in the reference feature data storage unit, a difference data storage unit different from data relating to differences between the objects is stored mutually similar,
A feature data extraction unit that extracts feature data indicating features of the subject from a target image in which the subject is represented;
Using the feature data and the plurality of reference feature data corresponding to the plurality of objects, a candidate object as a candidate for the subject expressed in the target image is selected from the plurality of objects. A candidate selector to select;
When a plurality of candidate objects are selected, a difference information providing unit that provides a user with difference information indicated by the difference data between the plurality of candidate objects;
An image processing apparatus comprising: The image processing apparatus according to claim 1, further comprising:
A plurality of detailed data corresponding to the plurality of objects, each detailed data indicating detailed information of the corresponding object, a detailed data storage unit storing the plurality of detailed data;
A detailed information providing unit that provides the user with the detailed information indicated by the detailed data corresponding to the designated candidate object when any of the plurality of candidate objects is designated by the user; ,
An image processing apparatus comprising: An image processing apparatus,
A plurality of reference feature data corresponding to a plurality of objects, each reference feature data indicating a feature of the corresponding object, and a reference feature data storage unit storing the plurality of reference feature data;
Of the plurality of objects corresponding to the plurality of reference data stored in the reference feature data storage unit, a difference data storage unit different from data relating to differences between the objects is stored mutually similar,
A subject identifying unit for identifying which of the plurality of objects is the subject represented in the target image;
With
The subject specifying unit
A feature data extraction unit that extracts feature data indicating the characteristics of the subject from the target image;
Using the feature data and the plurality of reference feature data corresponding to the plurality of objects, a candidate object as a candidate for the subject expressed in the target image is selected from the plurality of objects. A candidate selector to select;
A determination unit configured to determine any one of the plurality of candidate objects as the subject using the difference data between the plurality of candidate objects when the plurality of candidate objects are selected;
An image processing apparatus comprising: The image processing apparatus according to claim 3,
The determination unit
According to the difference data between the plurality of candidate objects, the user is designated a reference image in which a specific part of the same or same kind of subject as the subject represented in the target image is represented. An instruction unit for instructing;
A reference feature data extraction unit that extracts reference feature data indicating characteristics of the specific part of the subject from the reference image designated by the user;
With
The determination unit
An image processing apparatus that determines any one of the plurality of candidate objects as the subject using the reference feature data and the plurality of reference feature data corresponding to the plurality of candidate objects. The image processing apparatus according to claim 3, further comprising:
A plurality of reference detailed feature data corresponding to the plurality of objects, wherein each of the reference detail feature data stores a plurality of reference detailed feature data indicating detailed features of a specific part of the corresponding object. Provided with a detailed reference feature data storage unit,
The determination unit
A reference detail feature data selection unit that selects a plurality of reference detail feature data corresponding to the plurality of candidate objects based on the difference data between the plurality of candidate objects;
A detailed feature data extraction unit that extracts detailed feature data indicating detailed features of the specific part of the subject from the target image;
With
The determination unit
An image processing apparatus that determines any one of the plurality of candidate objects as the subject using the detailed feature data and the plurality of reference detailed feature data corresponding to the plurality of candidate objects. The image processing apparatus according to claim 3,
The determination unit
Based on the difference data between the plurality of candidate objects, comprising a setting unit for setting a weight for determining any of the plurality of candidate objects as the subject,
The determination unit
Image processing for determining any of the plurality of candidate objects as the subject using the feature data and the plurality of reference feature data corresponding to the plurality of candidate objects according to the weighting setting. apparatus. The image processing apparatus according to claim 6,
The determination unit
Using the feature data and the plurality of reference feature data corresponding to the plurality of candidate objects, a plurality of similarities are calculated, and the candidate object corresponding to the maximum similarity among the plurality of similarities Determine an object as the subject,
The determination unit
A plurality of elements calculated using a plurality of element data corresponding to a plurality of feature elements included in the feature data and a plurality of reference element data corresponding to the plurality of feature elements included in the reference feature data An image processing apparatus that calculates the respective similarities by applying the weights to similarities. The image processing apparatus according to claim 3, further comprising:
A plurality of detailed data corresponding to the plurality of objects, each detailed data indicating detailed information of the corresponding object, a detailed data storage unit storing the plurality of detailed data;
Details that provide the user with the detailed information indicated by the detailed data corresponding to the determined object candidate when any of the plurality of candidate objects is determined as the subject by the determining unit An information provider,
An image processing apparatus comprising:

Images