JP2007102290A - Image data management device, image data management method, and image data management program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To easily impart a keyword to image data even in equipment with insufficient UI. <P>SOLUTION: A distribution tendency acquired by taking statistics of condition parameters x<SB>0</SB>relating to sample image data 13b is compared with a condition parameter x acquired from the object image data 13b. Then, keywords with the largest number of condition items where the condition parameter x acquired from the object image data 13b is adapted to the distribution tendency are extracted, and some of those keywords are registered in object image data 13. Thus, it is possible to estimate the keywords whose condition parameters x have similar tendency as the candidates of registration, and to list up those keywords. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to an image data management device, an image data management method, and an image data management program, and in particular, image data for managing the image data by associating a keyword with a plurality of image data in which condition parameters are associated in advance. The present invention relates to a management apparatus, an image data management method, and an image data management program.

Conventionally, as this type of image data management apparatus, an apparatus that associates a desired keyword with image data and performs a search or the like using the keyword as a key is known (for example, see Patent Document 1). In this document, an image showing the characteristics of keyword candidates is displayed on the UI screen, and the keyword is determined by allowing the user to select an image.
According to such a configuration, it is possible to select a visually appropriate keyword and improve user convenience.
Japanese Patent Laid-Open No. 5-101117

However, the configuration described above is complicated because the user must select a keyword for each image data. Furthermore, there is a problem that a device that cannot provide a sufficient UI screen cannot display a list of images indicating the characteristics of keyword candidates.
SUMMARY An advantage of some aspects of the invention is that it provides an image data management device, an image data management method, and an image data management program capable of easily assigning keywords to image data even in devices having poor UI.

Means for Solving the Problems and Effects of the Invention

  In order to solve the above-mentioned problem, in the invention according to claim 1, a condition parameter is associated in advance with the image data. For example, in image data taken with a digital camera compliant with the EXIF standard, a header storing a plurality of condition parameters is attached to the net image data. In the EXIF standard, a plurality of condition parameters such as date information and shooting conditions are described by a tag method. It is possible to further associate a keyword with the image data associated with the condition parameter. Therefore, the image data can be organized using the keyword as a key. The keyword only needs to be associated with the image data main body, and may be attached as a header of the image data in the same manner as the condition parameter in the EXIF standard.

  The condition acquisition means acquires the condition parameter from the image data. For example, when the condition parameter is stored in the header of the image data, the header is analyzed. The estimation means holds the distribution tendency of the condition parameter of the sample image data statistically preliminarily for each of the keywords, and determines the compatibility between the distribution tendency and the condition parameter of the image data acquired by the condition acquisition means. get. Then, the keyword corresponding to the image data is estimated according to the suitability. Further, the correspondence unit associates the keyword estimated by the estimation unit with the image data. Accordingly, the keyword can be associated with the image data without the user inputting the keyword.

  The estimation means determines which keyword is based on the compatibility between the condition parameter of the image data acquired by the condition acquisition means and the distribution tendency of the condition parameter of the sample image data statistically preliminarily for each keyword. It is determined whether the image data is supported. For example, if the value of the condition parameter of the image data approximates the average value of the condition parameter of the sample image data to which a certain keyword corresponds, the compatibility is high, and the image data corresponds to the keyword. Can be determined. That is, the image data can be attributed to a keyword that approximates the condition parameter.

  Further, the condition parameter is not limited to a single one. In the invention according to claim 2, the condition parameter is associated with the image data for each of a plurality of condition items. That is, a plurality of condition items such as creation date, creation time, creation point, number of pixels, and the like are provided for a single image data, and the condition parameters indicating the contents of each condition item are stored. The estimation means obtains a matching value between the distribution tendency of the sample image data for each condition item and the condition parameter of the image data, and calculates an evaluation value obtained by combining the matching values. This evaluation value is calculated for each keyword, and the keyword corresponding to the image data is estimated according to the evaluation value. By estimating the keyword corresponding to the image data based on the evaluation value obtained by combining the adaptation values for each of the plurality of condition items, it is possible to perform comprehensive estimation over the plurality of condition items. It should be noted that various combining methods can be used for combining a plurality of the adaptation values, and the most common is linear combination.

  Further, as a preferred method for combining the adaptation values by the estimation means, the invention according to claim 3 weights the adaptation values for each condition item. Thereby, the above-mentioned condition items that should be emphasized and the above-mentioned condition items that may be neglected can be set, and the accuracy of estimation can be improved.

  As one of the simplest methods for calculating the evaluation value, in the invention according to claim 4, when the difference between the condition parameter and the central value of the condition parameter in the distribution tendency falls below a predetermined threshold value The fitness value is set to 1, and the fitness value is set to 0 when the difference exceeds a predetermined threshold. That is, when the difference between the condition parameter and the central value of the condition parameter in the distribution tendency falls below a predetermined threshold, the condition value is set to 1 for the condition item. On the other hand, if the difference exceeds a predetermined threshold value, the condition value is set to 0 for the condition item. By adding the matching values, the number of matching items among the condition items can be calculated as the evaluation value. Then, it is presumed that the keyword having many matching condition items corresponds to the image data. That is, the degree of conformity of the distribution tendency of the condition parameter is binary-determined, and it is estimated that the keyword having the largest number of conforming condition items corresponds to the image data. For the central value, an average value, a median value (median), a mode value, or the like can be applied. Of course, the present invention is not limited to simply adding the adaptation values, and addition may be performed while weighting each adaptation value.

  Furthermore, as a preferred example of the threshold value, in the invention according to claim 5, the threshold value is set based on the variation in the distribution tendency. For example, by using a constant multiple of the standard deviation as an index of the threshold, binary determination can be performed so that an arbitrary rejection rate is obtained in the normal distribution.

  The adaptation value is not limited to a binary value, and may be a multi-tone value. As a specific example, in the invention according to claim 6, the condition parameter and the center of the condition parameter in the distribution tendency are used. A value that increases as the difference from the value decreases is calculated as the fitness value. That is, the adaptation value is calculated for each of the condition items as a value that varies in multiple gradations or continuously according to the difference, and the evaluation value is calculated by adding the adaptation values. Therefore, it can be presumed that the keyword that increases the condition items with a small difference corresponds to the image data. Of course, the present invention is not limited to simply adding the adaptation values, and addition may be performed while weighting each adaptation value.

  On the other hand, the keyword may be determined by the user. In the invention according to claim 7, the estimating means lists a plurality of the keywords corresponding to the image data. Then, it is determined that the user is most suitable, and the keyword for which the selection instruction has been given is finally associated with the image data. That is, the estimation means narrows down the keywords to some extent, and finally the user determines the keywords. Thereby, it can prevent that the said keyword contrary to a user's will can be matched.

  Furthermore, in the invention according to claim 8 as a preferred example of the condition item, at least the creation time information at which the image data is created is included in the condition item, and the distribution tendency and the condition parameter relating to the creation time information are It is estimated that the matching keyword corresponds to the image data. For example, it is possible to accurately estimate the keyword for an event held every year or month.

  As another preferred example of the condition item, in the invention according to claim 9, at least the point information where the image data is created is included in the condition item, and the distribution tendency and the condition parameter regarding the point information are included. It is estimated that the matching keyword corresponds to the image data. For example, it is possible to accurately estimate the keyword for an event held at a certain place.

  Of course, the above invention can be realized not only by the apparatus but also by an image data management method as in claim 9, and image data management for executing processing according to the method as in claim 10. It can also be realized by a program. In addition, the idea of the invention is that the apparatus, method, and program according to the present invention may be implemented independently, or may be implemented together with other apparatuses, methods, and programs while being incorporated in a certain device. Is not limited to this, and includes various aspects, and can be changed as appropriate. For example, the image data management apparatus of the present invention may be incorporated in an image input / output device such as a digital still camera or a printer, or may be incorporated in another device such as an information home appliance.

  Further, it can be provided as a recording medium on which the program of the present invention is recorded. The recording medium for this program may be a magnetic recording medium, a magneto-optical recording medium, or any recording medium that will be developed in the future. In addition, the duplication stages such as the primary duplication product and the secondary duplication product are equivalent without any question. Further, even when a part is software and a part is realized by hardware, the idea of the invention is not completely different, and a part is stored on a recording medium and is appropriately changed as necessary. It may be in the form of being read. In addition, not all functions are necessarily realized by a single program, but may be realized by a plurality of programs. In this case, what is necessary is just to make each function implement | achieve in a some computer.

Here, embodiments of the present invention will be described in the following order.
(1) Computer configuration:
(2) Flow of distribution DB creation processing:
(3) Keyword registration process flow:
(4) Modification:

(1) Computer configuration:
FIG. 1 shows a schematic configuration of a computer as a specific image detection apparatus of the present invention. In FIG. 1, a computer 10 includes a CPU 11, a RAM 12, an HDD 13, a USB interface (I / F) 14, an input device interface (I / F) 15, and a video interface (I / F) 16 connected by an internal bus 10a. The HDD 13 stores various program data 13a, a plurality of image data 13b, 13b, 13b,... And a distribution database (DB) 13c. The CPU 11 reads out the program data 13a and executes processing based on the program data 13a while using the RAM 12 as a work area. A printer 20 and a digital still camera 30 are connected to the USB interface (I / F) 14, and a mouse 40 and a keyboard 50 are connected to the input device interface 15. Further, a display 60 is connected to the video interface (I / F) 16. Image data 13b, 13b, 13b... Is stored by transferring image data captured by the still camera 30 to the HDD 13.

  FIG. 2 shows a software configuration of a program executed on the computer 10. In the figure, an operating system (O / S) (not shown) is executed, and an image data management program P is executed as an application program on the O / S. The image data management program P provides various functions for organizing and managing a large number of image data 13b, 13b, 13b... Stored in the HDD 13. The image data management program P includes a condition acquisition unit P1, an estimation unit P2, a correspondence unit P4, an update unit P5, and a UI unit P6. The estimation unit P2 is compared with the matching value calculation unit P2a and the combination unit P2b. Part P2c.

  The UI unit P6 outputs necessary UI screens to the display 60 in response to requests from the other modules P1 to P5, and monitors the operation states of the mouse 40 and the keyboard 50. The UI unit P6 displays a UI screen for allowing the user to select any of image data 13b whose keyword to be described later is not registered, and accepts selection of the image data 13b. The condition acquisition unit P1 analyzes the header of the selected image data 13b and acquires a plurality of condition parameters stored in the header.

  FIG. 3 schematically shows a schematic configuration of the image data 13b. In the figure, the image data 13b includes a body B in which pixel data in which color signal values are stored for each of a plurality of pixels is stored, and a header H in which additional information regarding the image data 13b is stored. ing. For example, the body B stores pixel data in which colors are expressed by RGB color signal values or YCbCr color signal values. Further, the pixel data may be compressed and stored by a compression technique such as JPEG compression. In the header H, condition parameters are stored for each of a plurality of condition items.

  For example, data indicating October 10, 2005 is stored as a condition parameter in association with the condition item creation date (shooting date), and the condition item creation point (shooting point) is stored with the longitude and latitude as condition parameters. The data shown are stored in association with each other. Each condition parameter is attached with a tag conforming to a standard such as EXIF, and the condition acquisition unit P1 can analyze the tag to identify which condition item the condition parameter relates to. . The condition parameter is attached when image data is created when the digital still camera 30 performs shooting. The digital still camera 30 acquires the shooting time from a timer provided therein, and describes the acquired shooting time in the header H. In addition, the digital still camera 30 has a GPS [Global Positioning System] function, acquires the longitude and latitude of the shooting point at the time of shooting, and describes them in the header H. A mobile phone may be provided with a GPS function, and the digital still camera 30 built in such a mobile phone can acquire the longitude and latitude of the shooting point.

  The header H can store keywords in addition to the condition parameters. The condition parameter is described when the image data 13b is created (at the time of shooting), but the keyword is character information that can be registered later by the user. Since the user can arbitrarily register keywords, it is convenient that the user can organize the image data 13b for each keyword. When the user selects the desired image data 13b and inputs the desired keyword using the keyboard 50, the input keyword is written in the header H. Since the keyword is an arbitrary registration item of the user, the keyword may not be registered depending on the image data 13b. The matching value calculation unit P2a of the estimation unit P1 receives the condition parameter acquired by the condition acquisition unit P1, and calculates the matching value of each condition item for each keyword. When the fitness value calculation unit P2a calculates the fitness value, the distribution DB 13c prepared in advance in the HDD 13 is referred to.

  FIG. 4 shows the contents of the distribution DB 13c. In the figure, the distribution DB 13c stores an average value μ and a standard deviation σ for each condition item for each keyword. This average value μ and standard deviation σ correspond to the distribution tendency of the present invention. The distribution DB 13c is stored in advance in the HDD 13 and is created in advance by the update unit P5 statistically processing condition parameters for the image data 13b in which keywords are registered. The image data 13b stored in advance in the HDD 13 and registered with keywords is referred to as sample image data 13b.

Updating unit P5 classifies the sample image data 13b for each keyword, by category keyword, and calculates an average value μ and the standard deviation σ of the condition parameters x ₀ for each condition item describes the distribution DB 13 c. That is, the average value μ and the standard deviation σ are calculated for each condition item of each keyword. In order to avoid confusion, the condition parameter for the unregistered image data 13b for which the keyword acquired by the condition acquisition unit P1 is expressed as a condition parameter x, and the keyword for which the update unit P5 performs statistics is for the registered sample image data. shall notation condition parameters and condition parameters x _0. In the header H of FIG. 3, the creation date is collectively stored as a single condition item. However, in the distribution DB 13c, the yearly date, the date of the month, and the day of the week are individual condition items. It has been statistics as. The date of the year and the date of the month can be obtained from the creation date of the header H. On the other hand, the day of the week cannot be obtained directly from the date of creation of the header H, but the day of the week can be obtained indirectly by providing the update unit P5 with calendar information. Of course, it is also possible to acquire the 6th day of Daan, Tomohiki, etc. from the calendar information and manage that 6th day as one of the condition items.

The adaptation value calculation unit P2a calculates threshold values T1 and T2 for the condition parameter x. The threshold values T1 and T2 are calculated by the following formulas (1) and (2), respectively.
T1 = μ−2σ (1)
T2 = μ + 2σ (2)
Then, it is determined whether the condition parameter x acquired by the condition acquisition unit P1 is larger than the threshold T1 and smaller than the threshold T2 (T1 <x <T2). Then, when the condition parameter x is larger than the threshold value T1 and smaller than the threshold value T2, it is specified that the fitness value S is 1 as conforming. On the other hand, when the condition parameter x is equal to or less than the threshold value T1 or equal to or greater than the threshold value T2, it is determined that the fitness value S is 0 because it does not match. The average value μ and the standard deviation σ of the condition parameter x ₀ for each condition item of each as described above keyword is calculated, respectively thresholds T1, T2 and adaptation value S for each condition item of each keyword Will be calculated.

  For example, a matching value of 0 or 1 is specified for the keyword “athletic day” for each condition item of annual day, month day, weekday, time, latitude, longitude, etc., and the keyword “cherry blossom viewing” A conforming value of 0 or 1 is specified for each condition item of day, month day, week day, time, latitude, longitude, and so on. The combining unit P2b calculates the evaluation value E by adding all the matching values S of the keywords. Therefore, the evaluation value E is calculated by the number of existing keywords. The comparison unit P2c compares the evaluation values E of the keywords, and extracts the top three keywords that are larger as candidate keywords. That is, the candidate keyword extracted here is estimated to be a keyword corresponding to the image data 13b from which the condition acquisition unit P1 has acquired the condition parameter x.

  The corresponding unit P4 acquires candidate keywords and generates a UI screen for listing the candidate keywords. The generated UI screen is displayed on the display 60 by the UI unit P6, and a selection operation for confirming any of the candidate keywords is accepted from the user via the mouse 40 or the keyboard 50. The corresponding unit P4 describes the candidate keyword selected by the user in the header H of the image data 13b. Thereby, a keyword can be newly registered with respect to the image data 13b in which the keyword has not been registered. If a keyword is registered, the image data 13b can be handled as sample image data 13b, and the update unit P5 can update the distribution DB 13c using the keyword and the condition parameter of the sample image data 13b.

(2) Flow of distribution DB creation processing:
FIG. 5 shows the flow of the distribution DB creation process performed by the update unit P5. In step S100, the image data 13b stored in the HDD 13 with the keyword already registered is extracted as sample image data 13b. In step S110, keywords registered in the sample image data are listed and one of them is selected. For example, first, the keyword “athletic meet” is selected. In step S120, sample image data 13b in which the keyword “athletic meet” is registered is extracted.

In step S130, it acquires the condition parameter x ₀ for each condition item from the header H of the sample image data 13b extracted in step S120, calculates an average value μ and the standard deviation σ for each condition item. Specifically, the average value μ and standard deviation σ for the day of the year are calculated, then the average value μ and standard deviation σ for the day of the month are calculated, and the average value μ and standard deviation σ for the weekday are further calculated. The calculation process is performed in the same manner for all condition items. Thus, the average value μ and standard deviation σ of each condition item can be calculated for the keyword “athletic meet”. The calculated average value μ and standard deviation σ are described in the distribution DB 13c of the HDD 13 in step S140.

  When the average value μ and the standard deviation σ of all the condition items for the keyword “athletic meet” selected first are finished, whether or not the calculation of the average value μ and the standard deviation σ has been completed for all the keywords in step S150. Determined. When the keyword “athletic meet” is selected for the first time, the calculation of the average value μ and standard deviation σ for the keywords “cherry blossom viewing”, “XX meeting”, etc. is incomplete, so the process returns to step S110 and the next keyword Is selected. By repeatedly performing the above processing, the average value μ and standard deviation σ of each condition item can be calculated for all keywords registered in advance.

  When the calculation of the average value μ and the standard deviation σ of each condition item is completed for all keywords, the creation of the distribution DB 13c is completed. The distribution DB creation processing described above may be performed at a stage where a certain number of sample image data 13b is stored in the HDD 13. For example, image data 13b in which keywords are registered is newly stored or existing in the HDD 13. The image data 13b may be performed when a keyword is newly registered by a keyword registration process described later. By doing in this way, distribution DB13c can always be updated to the newest state.

(3) Keyword registration process flow:
FIG. 6 shows the flow of the keyword registration process. Before the keyword registration process is executed, the distribution DB 13c needs to be created in advance by the distribution DB creation process. In step S <b> 200, the UI unit P <b> 6 accepts an operation for selecting any one of the image data 13 b whose keyword is not registered among the image data 13 b stored in the HDD 13. That is, the user performs a selection operation for registering a keyword for any one of the image data 13b in which the keyword is not registered. In step S210, the condition acquisition unit P1 acquires the condition parameter x by analyzing the header H of the image data 13b selected in step S200 (hereinafter referred to as target image data 13b). Since the condition parameter x is stored for each condition item as described above, the condition parameters x are acquired as many as the number of condition items.

  In step S220, the matching value calculation unit P2a selects one of the keywords described in the distribution DB 13c. For example, the following description will be given assuming that the keyword “athletic meet” is selected first. In step S230, the average value μ and the standard deviation σ of each condition item for the keyword “athletic meeting” are acquired from the distribution DB 13c, and threshold values T1 and T2 are calculated for each of them by the above formulas (1) and (2). Then, a matching value S of 1 or 0 is calculated for each condition item by comparing the condition parameter x of each condition item with the threshold values T1 and T2 of each condition item. Specifically, it is determined whether or not the condition parameter x is larger than the threshold value T1 and smaller than the threshold value T2. If the condition parameter x is larger than the threshold value T1 and smaller than the threshold value T2, the condition is met. Suppose that the fitness value S is 1. On the other hand, when the condition parameter x is equal to or less than the threshold value T1 or equal to or greater than the threshold value T2, it is assumed that the matching value S is 0 because it does not match.

上記式（３）によって、キーワード“運動会”についての条件項目“月間日”の条件パラメータｘ₀の分布傾向が正規分布であると仮定した場合の確率密度ｆ（ｘ）を算出することができる。 FIG. 7 is a graph illustrating the relationship between the threshold values T1 and T2 and the adaptation value S. In the figure, the abscissa indicates the condition parameters x and x ₀ , the ordinate indicates the probability density f (x), and the threshold values T 1 and T 2 of the condition item “monthly day” and the fitness value S for the keyword “athletic day” are shown. It shall be shown. The probability density f (x) can be expressed by the following formula (3).

The probability density f (x) when it is assumed that the distribution tendency of the condition parameter x ₀ of the condition item “monthly day” for the keyword “athletic day” is a normal distribution can be calculated by the above equation (3).

In the figure, the area surrounded by the threshold values T1 and T2 is shaded, and the integrated value of the threshold values T1 to T2 of the probability density f (x) that means the area of the hatched area is about 95%. Yes. This is about 95% of the sample image data 13b belonging to the keyword "athletic meet" means that a condition parameter x ₀ of condition item "Monthly Review" is the value of the threshold T1 to T2. In other words, if the condition parameter x of the condition item “monthly day” is the value of the thresholds T1 to T2, the target image data 13b from which the condition parameter x is acquired can belong to the keyword “athletic meet”. It can be judged that the nature is high. Accordingly, if the condition parameter x of the condition item “monthly day” has the value of the thresholds T1 to T2, the condition parameter x of the condition item “monthly day” is determined to be compatible with the keyword “athletic day” and the fitness value S Is set to 1.

  On the other hand, if the condition parameter x of the condition item “monthly day” is not the value of the thresholds T1 to T2, the target image data 13b from which the condition parameter x is acquired is very likely to belong to the keyword “athletic meet”. Can be judged low. Therefore, if the condition parameter x of the condition item “monthly day” does not have the values of the thresholds T1 to T2, the condition parameter x of the condition item “monthly day” is determined to be incompatible with the keyword “athletic meet” and the conformance value S Is set to 0. For example, if the target image data 13b is created (photographed) on October 9 (Sunday), the condition parameters x (x = October) related to the condition items “annual day”, “monthly day”, and “weekly day”. 9th, 9th, Sunday) will be suitable.

  When the matching value S of the keyword “athletic meet” is specified for each condition item as described above, the evaluation value E is calculated by combining the matching values S for each condition item in step S240. Specifically, the combining unit P2b simply adds the fitness value S for each condition item. Thereby, the number of condition items determined to be compatible can be counted. That is, it can be determined that the larger the evaluation value E, the more condition items that match the condition parameter x of each condition item acquired from the target image data 13b.

  In step S250, it is determined whether or not the evaluation value E has been calculated for all keywords. If not, the process returns to step S220, and the next keyword is selected. For example, after selecting the keyword “athletic meet”, the keyword “cherry-blossom viewing” is selected. Steps S230 and S240 are executed for the next selected keyword, and an evaluation value E obtained by adding the adaptation value S for each condition item and the adaptation value S is calculated. By repeating the above processing, the evaluation value E can be calculated for all keywords.

  FIG. 8 is a table showing the results of calculating the evaluation value E for all keywords. In the same figure, a matching value S of 0 or 1 is specified for each condition item of each keyword, and the evaluation value E is calculated by summing the matching values S for each keyword. When it is confirmed in step S250 that the evaluation values E have been calculated for all the keywords, the comparison unit P2c compares the evaluation values E of the keywords in step S260, and the keywords having the same evaluation value E are ranked from the top. Extract one. That is, it is estimated that three keywords having a large evaluation value E are keywords corresponding to the target image data 13b. In the example shown in FIG. 8, the keywords “athletic meet”, “cherry blossom viewing”, and “xx practice” are extracted. In step S270, the corresponding unit P4 acquires the three keywords extracted by the comparison unit P2c in step S260, and the corresponding unit P4 generates a UI screen on which these keywords are listed, and outputs them to the UI unit P6. To do.

  FIG. 9 shows an example of the UI screen displayed in step S270. In the figure, the keywords “athletic meeting”, “cherry blossom viewing”, and “practice of xx” extracted by the comparison unit P2c in step S260 are listed as candidate keywords, and any one of these is listed as a mouse 40 or a keyboard 50. There is provided a check box that can be instructed to select. Corresponding unit P4 accepts a candidate keyword selection instruction via UI unit P6. The keyword selected in step S280 is associated with the target image data 13b. Specifically, the selected keyword is added to the header H of the image data 13b. When the keyword association is completed as described above, the keyword registration process is completed.

  When the keyword registration process is completed, a search or the like can be performed based on the registered keywords. For example, when a search is performed by specifying the keyword “athletic meet”, the image management program P analyzes the header H of each image data 13b, 13b, 13b..., And the image data 13b, 13b, 13b.・ ・ Get the keyword. Then, by extracting only the image data 13b, 13b, 13b... Whose keyword is “athletic meet”, it is possible to realize a search for image data using the keyword as a key. In addition, by executing the above-described distribution DB creation process when the keyword registration process is completed, the distribution DB 13c reflecting the registered keyword can be updated.

In the present embodiment, the distribution tendency obtained by statistically analyzing the condition parameter x ₀ regarding the sample image data 13b is compared with the condition parameter x obtained from the target image data 13b. Then, a keyword having a large number of condition items whose condition parameter x acquired from the target image data 13 b matches the distribution tendency is extracted, and any of them is registered in the target image data 13. By doing in this way, it is possible to estimate and list keywords that are likely to be similar in condition parameter x as candidates to be registered. For example, if “annual day” as a condition item has a similar tendency, it can be determined that the keyword of an event performed in an annual cycle is suitable.

  Similarly, if “Monthly Day” as a condition item has a similar tendency, it can be determined that the keyword of the event that is performed in the monthly cycle is applicable, and if “Weekly Day” tends to be similar, the keyword of the weekly event It can be judged that it fits. As described above, each keyword tends to have a certain tendency at the time of creation (photographing). Therefore, an appropriate keyword can be estimated by using creation time information as a condition item.

On the other hand, information on the creation (photographing) point is also effective information for estimating the corresponding keyword. For example, in the image data 13b taken when traveling, there is a specific tendency in the created longitude and created latitude. Therefore, an appropriate keyword can be estimated accurately by setting the creation point information as a condition item. Since the matching value S is calculated for each of these condition items, and the keyword corresponding to the target image data 13b is finally estimated by the evaluation value E obtained by adding the matching value S, a comprehensive value is obtained from all the condition items. It is possible to extract a keyword having a similar distribution tendency of the condition parameter x ₀ as shown in FIG.

(4) Modification:
In the above-described embodiment, a comprehensive evaluation value E is calculated for each condition item by specifying a matching value S of 0 or 1 in each condition item and simply adding the matching value S. However, the importance may differ depending on the condition item. For example, when a regular event is often performed at a plurality of places, the importance of the creation time information is higher than the creation point information of longitude and latitude. In such a case, it is preferable to perform weighting according to the importance when adding the conforming value S of each condition item. In the above example, since the longitude and latitude of the condition items can be neglected, the addition may be performed after the fitness value S relating to longitude and latitude is multiplied by a coefficient of 0.5. As a result, the weight of the condition item relating to the creation time is relatively increased, and the evaluation value E that places importance on the creation time can be obtained. In the above-described embodiment, the fitness value S is determined to be either 0 or 1. However, the fitness value S having an intermediate value may be calculated. For example, the fitness value S may be calculated by the following formula (4).

FIG. 10 is a graph showing the relationship between the condition parameter x and the adaptation value S according to the above equation (4). In the figure, the fitness value S has the same tendency as the probability density in the normal distribution. The smaller the difference between the condition parameter x and the average value μ, the larger the value, and conversely, the difference between the condition parameter x and the average value μ. The larger the difference, the smaller the value. By calculating such an adaptation value S, it is possible to reflect the degree to which the condition parameter x is close to the average value μ in the adaptation value S while taking into account variations in the condition parameter x ₀ of the sample image data 13b. Even when such matching values S are combined, a comprehensive evaluation value can be obtained for each condition item, and weighting according to the importance of each condition item can be performed when combining.

  There may be a relationship between each condition item. For example, when the image data 13b is created not in Japan but in a foreign country, the possibility of an annual event being reduced is the same as in Japan. Accordingly, the distribution DB 13c in which the keywords are classified for each country is created, the shooting country of the target image data 13b is first determined, and the keywords for evaluating the condition parameter x are narrowed down according to the determination result. Good. By doing so, it is possible to prevent a keyword based on a Japanese annual event from being estimated for the target image data 13b taken in a foreign country, and to estimate a keyword corresponding to the foreign annual event. be able to.

  In the above-described embodiment, an example in which a large number of image data 13b and the distribution DB 13c are stored in the computer 10 is illustrated. However, the image data 13b and the distribution DB 13c are not necessarily stored in the same device. Absent. Since the distribution DB 13c is data storing the average value μ and the standard deviation σ for each keyword, the amount of data can be extremely small. Accordingly, the distribution DB 13c can be stored in a printer or digital camera having a small storage capacity, and keyword estimation based on the distribution DB 13c can be performed in the printer or digital camera. In this case, many image data 13b itself is separately stored in a computer, a photo server, a photo storage, or the like. The threshold values T1 and T2, the adaptation value S, and the evaluation value E can all be obtained by simple calculations, and keyword estimation can be realized even with a small-scale computing device. Therefore, the present invention can be easily applied to a printer, a digital camera, or the like.

  In addition, since it is not necessary to display a UI for directly inputting a keyword or visually selecting a keyword for each image data 13b, a keyword or a digital camera or the like having a poor UI such as a screen / input device may be used. Registration can be done easily. However, since it is necessary to update the distribution DB 13c, it is desirable that a printer, a digital camera, or the like can access a computer, a photo server, a photo storage, or the like that stores a large number of image data 13b. In the above-described embodiment, the user finally selects the estimated keyword. However, in a printer or digital camera having a poor UI, the keyword having the largest evaluation value S is used as it is as the target image data 13b. You may make it match | combine.

It is a hardware block diagram of a computer. It is a software block diagram of a program. It is a schematic block diagram of image data. It is a table | surface which shows distribution DB. It is a flowchart which shows the flow of distribution DB creation processing. It is a flowchart which shows the flow of a keyword registration process. It is a graph explaining threshold value T1, T2. 5 is a table showing calculation results of a conformance value S and an evaluation value E. It is a figure which shows an example of UI screen. It is a graph which shows the relationship between the conformity value S and the condition parameter x.

Explanation of symbols

DESCRIPTION OF SYMBOLS 10 ... Computer, 10a ... Bus, 11 ... CPU, 12 ... RAM, 13 ... HDD, 13a ... Program data, 13b ... Image data, 13c ... Distribution DB14 ... USB I / F, 15 ... Input apparatus I / F, 16 ... Video I / F, 20 ... printer, 40 ... mouse, 50 ... keyboard, 60 ... display, P ... image data management program, P1 ... condition acquisition unit, P2 ... estimation unit, P2a ... adaptation value calculation unit, P2b ... combination unit, P2c ... comparison unit, P4 ... corresponding unit, P5 ... update unit, P6 ... UI unit

Claims (11)

In an image data management apparatus that associates a keyword with a plurality of image data in which condition parameters are associated in advance,
Condition acquisition means for acquiring the condition parameter from the image data;
Estimating means for estimating the keyword corresponding to the image data based on the compatibility between the acquired condition parameter of the acquired image data and the distribution tendency of the condition parameter of the sample image data previously statistically calculated for each keyword When,
An image data management apparatus comprising: a correspondence unit that associates the estimated keyword with the image data. The image data is associated with the condition parameter for each of a plurality of condition items,
The estimation means is
For each keyword, while calculating an evaluation value that combines the distribution tendency of the sample image data for each condition item and the matching value of the condition parameter of the image data,
The image data management apparatus according to claim 1, wherein the keyword corresponding to the image data is estimated based on the evaluation value. The estimation means is
3. The image data management apparatus according to claim 2, wherein weighting is performed for each of the condition items when combining the adaptation values. The estimation means is
When the difference between the condition parameter and the central value of the condition parameter in the distribution tendency is below a predetermined threshold, the fitness value is 1, and when the difference is above the threshold, the fitness value is 0. With
4. The image data management apparatus according to claim 2, wherein the keyword having a large evaluation value obtained by adding the matching values is estimated to correspond to the image data.   The image data management apparatus according to claim 4, wherein the threshold is set based on variation in the distribution tendency. The estimation means is
While calculating the evaluation value by adding the adaptation value, the value becomes larger as the difference between the condition parameter and the central value of the condition parameter in the distribution tendency is smaller,
The image data management apparatus according to claim 2, wherein the keyword having a large evaluation value obtained by adding the matching values is estimated to correspond to the image data. The above countermeasures are
2. The plurality of keywords estimated by the estimating means to correspond to the image data, and the keyword for which a selection instruction has been received is associated with the image data. 6. The image data management device according to any one of 6.   8. The image data management apparatus according to claim 2, wherein at least the condition item includes creation time information when the image data is created.   9. The image data management apparatus according to claim 2, wherein at least the condition item includes point information where the image data is created. In an image data management method for associating a keyword with a plurality of image data in which condition parameters are associated in advance,
A condition acquisition step of acquiring the condition parameter from the image data;
An estimation step of estimating the keyword corresponding to the image data based on the suitability between the condition parameter of the acquired image data and the distribution tendency of the condition parameter of the sample image data previously statistically calculated for each keyword When,
And a corresponding step of associating the estimated keyword with the image data. In an image data management program for causing a computer to execute a function of associating a keyword with a plurality of image data in which condition parameters are associated in advance.
A condition acquisition function for acquiring the condition parameter from the image data;
An estimation function for estimating the keyword corresponding to the image data based on the compatibility between the condition parameter of the acquired image data and the distribution tendency of the condition parameter of the sample image data previously statistically calculated for each keyword When,
An image data management program that causes a computer to execute a corresponding function for associating the estimated keyword with the image data.

Images