JP2007026316A - Image management device, image-managing computer program and recording medium recording the same - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an image management device capable of reducing trouble of information addition to individual image data, and allowing accurate retrieval. <P>SOLUTION: A plurality of image data including a person image are stored in a photograph database 20. A person image detection part 31 extracts a person image (e.g. face image) area inside each the image data. Image is recognized by a person image recognition processing part 32 in an image of the extracted area, and a recognition result is stored in the photograph database 20. Correction input of the recognition result by a user is received by a correction input reception part 34. According thereto, a recognition result correction processing part 35 corrects the recognition result stored in the photograph database 20. The the image data is retrieved by an image retrieval processing part 39 according to a retrieval condition received through a retrieval condition input reception part 38. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to an image management apparatus for managing a plurality of pieces of image data including a person image, a computer program for causing a computer to function as such an image management apparatus, and a computer readable recording such a program The present invention relates to a recording medium.

Conventional techniques for searching for photographic data showing a specific person from a large number of photographic data are disclosed in Patent Documents 1 to 3 below.
Patent Documents 1 and 2 disclose a technique in which names, dates, and other information are added in advance to individual photos, and specific photos are searched based on this information. However, since the task of adding information such as a name to each photo is a precondition for the search, it takes a great deal of time to prepare for using the search function.

On the other hand, Patent Document 3 discloses a technique for automatically attaching information to individual photographs using an image recognition technique and retrieving a target photograph based on the information. However, since it is not easy to automatically recognize a person in various poses in a plurality of photos, the reliability of information automatically assigned is low. Therefore, there is a problem that it is difficult to obtain an accurate search result.
Japanese Patent Laid-Open No. 11-066089 JP 2001-084274 A Japanese Patent Laid-Open No. 11-250071

SUMMARY OF THE INVENTION An object of the present invention is to provide an image management apparatus that can reduce the effort of adding information to individual image data and can perform an accurate search.
Another object of the present invention is to provide a computer program for causing a computer to function as such an image management apparatus.
Still another object of the present invention is to provide a recording medium on which such a computer program is recorded.

  In order to achieve the above object, the invention according to claim 1 is an image data storage means for storing a plurality of pieces of image data including a person image (for example, a face image), and is stored in the image data storage means. A person image recognizing means for recognizing a person image in the individual image data, a recognition result storing means for storing the recognition result by the person image recognizing means in association with each image data, and the person image recognizing means. A correction input receiving means for receiving a correction input for correcting the recognition result, and a recognition result correction for correcting the recognition result stored in the recognition result storage means in accordance with the correction input received by the correction input receiving means. And an image management apparatus.

  According to this configuration, human images in a plurality of pieces of image data are recognized by the human image recognition unit and stored in the recognition result storage unit. A correction input for this recognition result is received by the correction input receiving means. When the correction input is received by the correction input receiving means, the recognition result correcting means corrects the recognition result according to the correction input. Thus, accurate information can be added to individual image data with a small number of steps by a configuration in which the person image recognition process by the person image recognition means is supplemented by correction input by the operator. By using the information added in this way, it is possible to accurately search image data.

The person image recognition means may perform person recognition processing based on an arbitrary feature that can identify a person in an image. Examples of such features include a face image of a person in the image, a whole image of the person, a half-body image, and clothes (color of clothes, etc.).
The correction input accepting unit may accept a selection input of any of a plurality of recognition candidates (candidate persons). Thereby, the recognition result can be easily corrected.

Moreover, it is preferable that the said correction input reception means further includes the new registration means which receives the input of the new recognition candidate other than the recognition candidate by the said person image recognition means. Accordingly, appropriate information can be added to a person image that cannot be recognized by the person image recognition means.
The correction input accepting unit preferably includes a registration unit without additional information for registering “no additional information” for a person image in the image (preferably, an arbitrary person image selected by the user). For example, by registering “no additional information” for a stranger person image of a stranger, a necessary person image can be searched more accurately.

  The invention according to claim 2 further includes display signal generation means for generating a display signal for displaying the recognition result by the person image recognition means on the display device together with the image data stored in the image data storage means. The image management apparatus according to claim 1. According to this configuration, the operator can correct the recognition result while visually confirming the recognition result.

When the correction input accepting unit accepts a selection input from among a plurality of recognition candidates, the display signal generating unit generates a display signal for displaying the plurality of recognition candidates. It is preferable that
The invention according to claim 3 is a dictionary storing means for storing dictionary data for recognizing a person image, and a dictionary for correcting dictionary data in the dictionary storing means based on the recognition result corrected by the recognition result correcting means. The human image recognition means further comprises a correction means for recognizing a human image in the image data based on dictionary data stored in the dictionary storage means. 2. The image management apparatus according to 2.

According to this configuration, when the recognition result is corrected, dictionary data for recognizing the person image is corrected. Thereby, every time the operator performs a correction operation, the accuracy of the dictionary data is increased, and the accuracy of the human image recognition is increased. As a result, information can be added to the image data more efficiently.
When the correction input receiving means includes the new registration means, it is preferable that the dictionary correction means adds newly registered new recognition candidate data to the dictionary data. Thereby, the recognition candidates which can recognize an image can be increased. In this way, recognition candidate data can be accumulated in the dictionary data by a new registration operation, so that human image recognition accuracy can be improved and correction input work by the user can be reduced.

Dictionary data is a set of feature data for each recognition candidate. The feature amount data may be recognition candidate image data, or may represent a predetermined feature amount extracted from such image data.
The invention according to claim 4 further includes a recognition result determination unit for determining a recognition result stored in the recognition result storage unit, wherein the dictionary correction unit is configured to use the dictionary based on the determined recognition result. 4. The image management apparatus according to claim 3, further comprising means for correcting dictionary data in the storage means.

  According to this configuration, the dictionary data is corrected by confirming the recognition result. More specifically, when the recognition result for a certain person image is determined, the dictionary data is corrected so as to increase the probability that the same recognition result is obtained for the person image. More specifically, the feature amount data of the person image for which the recognition result is confirmed may be added to the dictionary storage means as dictionary data corresponding to the recognition candidate of the recognition result. By performing person image recognition processing using such dictionary data, feature amount data in various poses and lighting conditions is accumulated for each recognition candidate, enabling highly accurate person recognition. Become.

The recognition result determination means may write confirmation completion data indicating that the recognition result has been confirmed by the user into a predetermined confirmation completion data storage means. For example, the confirmation completion data may be a “confirmation flag” indicating confirmation / unconfirmation of the recognition result.
The recognition result determining means may include means for automatically determining the recognition result corrected by the recognition result correcting means. In this case, by correcting the dictionary data based on the confirmed recognition result, the purpose of correcting the dictionary data based on the recognition result corrected by the recognition result correcting means can be achieved as a result.

Further, a confirmation input accepting unit that accepts a confirmation input for confirming the recognition result may be further provided. In this case, it is preferable that the recognition result confirmation unit includes a unit that confirms the recognition result stored in the recognition result storage unit in response to the confirmation input received by the confirmation input reception unit.
It is preferable that the person image recognizing unit excludes a human image whose recognition result is confirmed by the recognition result confirming unit from a recognition process target. Thereby, it is possible to prevent the correct recognition result from being corrected.

The invention according to claim 5 further includes a search means for searching image data based on the recognition result stored in the recognition result storage means. It is a management device.
According to this configuration, the image data can be searched based on the accurate information added to the image data, so that the image data including the target person image can be searched with high accuracy.

  According to a sixth aspect of the present invention, the person image recognition means includes likelihood calculation means for calculating the likelihood that the person image is a recognition candidate image for each of a plurality of recognition candidates, and image data including a plurality of person images. And a priority setting unit that determines the priority of the recognition candidate according to the likelihood obtained by the likelihood calculating unit when the maximum likelihood recognition candidate overlaps between two or more person images. An image management apparatus according to any one of claims 1 to 5.

  According to this configuration, the person image recognition means can calculate the likelihood for each recognition candidate and can determine the recognition result based on the order based on the likelihood. On the other hand, as a general rule, images of a real scene such as a photographed image do not have duplicate images of the same person. Therefore, when the recognition candidates of the maximum likelihood overlap for two or more person images in one image data, the maximum likelihoods are compared with each other. The recognition candidate is given priority for the person image having the largest maximum likelihood, and the recognition candidates for the other person images are changed to the recognition candidates having the second or lower likelihood. Thus, since a reasonable recognition result can be obtained, the number of steps for correcting the recognition result by the user can be reduced.

  The invention according to claim 7 further includes a recognition result confirmation means for confirming a recognition result stored in the recognition result storage means, wherein the correction input accepting means provides a plurality of recognition candidates to the user. A recognition candidate in the image data when the recognition result of one of the person images is determined by the recognition result determination means in the image data including a plurality of person images. The recognition candidate excluding means for excluding the confirmed recognition result from the recognition candidate options presented by the selection candidate presenting means for a human image whose result is not yet confirmed, further comprising: The image management apparatus according to claim 1.

In this configuration, the user can correct the recognition result by selecting one of the recognition candidate options presented by the selection candidate presenting unit.
In principle, there is no duplicate image of the same person in an image of a real scene such as a photographed image. Therefore, when the recognition result is confirmed for any person image in the image data, the recognition candidate for the confirmed recognition result is excluded from the choices of recognition candidates for the unconfirmed person image. As a result, only reasonable selection candidates are presented by the selection candidate presenting means, so that the user can easily select recognition candidates and input errors can be reduced. Thereby, it is possible to more efficiently add information to the image data.

The invention according to claim 8 is characterized in that the person image recognition means generates index image specifying data for specifying index image data corresponding to the recognition result as the recognition result data. The image management apparatus according to any one of claims 7 to 9.
According to this configuration, index image specifying data is generated as recognition result data. Thereby, the index image data specified by the index image specifying data can be used as additional information for the person image in the image data. In this way, information (index image) in a format that allows more intuitive grasping can be added to the image data.

  In this case, it is preferable that an index image data storage unit that stores in advance an index image corresponding to the index image specifying data is provided. Further, in the case of the configuration of the invention of claim 2, the display signal generation means reads index image data corresponding to the index image specifying data as the recognition result from the index image data storage means, and the index image data is recognized as the recognition result. It is preferable to generate a display signal for display on the display means.

The index image data may be data representing an image that can identify a person candidate. For example, the index image data may be a representative face image of a person candidate, an icon, a character, or another simple display image. Good.
The invention according to claim 9 further includes a recognition result determination means for determining a recognition result stored in the recognition result storage means, and the image data stored in the image data storage means are ordered. The person image recognition means executes person image recognition processing for a plurality of image data stored in the image data storage means according to the ordering, and the recognition result confirmed by the recognition result confirmation means 9. A recognition result automatic confirming unit for automatically confirming recognition results regarding image data of a rank prior to corresponding image data (all unconfirmed recognition results are confirmed). Image management apparatus.

  For example, the operator browses the recognition results in the order of the image data and performs necessary corrections. In this case, it is not necessary to confirm the recognition result regarding the intermediate image data one by one. That is, the recognition result regarding the image data in the order before the image data is automatically determined by determining the recognition result regarding the image data in the subsequent order. In this way, the operator's input effort can be significantly reduced.

The recognition result determination means may write confirmation completion data indicating that the recognition result has been confirmed by the user into a predetermined confirmation completion data storage means.
The invention according to claim 10 is a computer program for operating a computer as an image management apparatus for managing a plurality of image data including a person image, wherein the computer includes a plurality of images including a person image. Image data storage means for storing data, and recognition result storage means for storing a recognition result of a person image, wherein the computer program stores each computer stored in the image data storage means. Person image recognition means for recognizing a person image in image data, means for writing the recognition result by the person image recognition means in association with each image data into the recognition result storage means, and correcting the recognition result by the person image recognition means Correction input receiving means for receiving correction input for correction, and correction input receiving means An image management computer program that functions as a recognition result correction unit that corrects a recognition result stored in the recognition result storage unit according to a received correction input.

By executing this computer program on a computer, the image management apparatus of claim 1 can be configured.
Of course, the various modifications described for the image management apparatus can be applied to this computer program.
The invention described in claim 11 is a computer-readable recording medium on which the computer program according to claim 10 is recorded.

By making a computer read the computer program recorded on this recording medium and making it run, this computer can be functioned as an image management apparatus as mentioned above.
The computer program may be recorded on the recording medium in any format, for example, may be recorded in a compressed format. As the recording medium, any recording format including a magnetic recording medium, an optical recording medium, and a magneto-optical recording medium can be applied.

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings.
FIG. 1 is a block diagram for explaining a basic configuration of a computer operable as an image management apparatus according to an embodiment of the present invention. The computer includes a computer main body 1, an input operation unit 2 connected to the computer main body 1, and a display device (display) 3 also connected to the computer main body 1, and has a form as a personal computer. Yes. The input operation unit 2 and the display device 3 provide a man-machine interface for the computer main body 1. The user performs a necessary input operation from the input operation unit 2 while confirming the display on the display device 3.

Examples of the input operation unit 2 include a keyboard, a pointing device (for example, a mouse or a track pad), and a touch panel (arranged on the display screen of the display device 3). The display device 3 includes, for example, a CRT, a liquid crystal display panel, or other two-dimensional display device.
The computer main body 1 includes a CPU 5, a ROM 6, a RAM 7, a hard disk drive (HDD) 8 as an external storage device, a CD-ROM drive 9, a display controller 10, an input controller 11, a communication interface 12 (USB interface or LAN interface), and A memory card slot 13 is provided, and these are connected to the bus 14.

  The CD-ROM drive 9 can be loaded with a CD-ROM 15 as a recording medium on which a computer program (digital photo album program; image management program) for operating this computer as an image management apparatus is recorded. When the CD-ROM 15 is loaded into the CD-ROM drive 9, the computer program recorded there can be read. When the user performs an installation operation by operating the input operation unit 2, the computer program is expanded in the storage area of the hard disk drive 8. By executing this computer program, the computer can be operated as an image management apparatus. That is, the CPU 5 executes the computer program in cooperation with the ROM 6, the RAM 7, and other components in the computer main body 1.

Such a computer program need not be provided in the form of a CD-ROM, but may be provided in the form of a magneto-optical disk or any other type of recording medium. Further, a computer program may be acquired via a communication line connected to the communication interface 12 and stored in a storage area of the hard disk drive 8.
The display controller 10 is for controlling display on the display device 3 based on a command from the CPU 5. The input unit controller 11 is for receiving an input from the input operation unit 2 and transmitting it to the CPU 5.

  By executing the computer program, the CPU 5 stores the photo database 20 (image data storage means, recognition result storage means and confirmation completion data storage means) and recognition dictionary storage unit 21 (dictionary storage means) in the storage area of the hard disk drive 8. ). The photo database 20 can store data of a plurality of photo images. That is, the photo database 20 stores photo data including image data representing a photo image and additional data related to the image data. The recognition dictionary storage unit 21 stores dictionary data for recognizing a person in the image data. This dictionary data includes a group of feature amount data of face images of a plurality of persons who are recognition candidates.

  The feature amount data is collation data to be collated with image data to be recognized, and is data representing feature amounts useful for distinguishing recognition candidate face images from other people's face images. Specifically, the feature amount data may be recognition candidate face image data itself, or may be data extracted by performing predetermined processing on the face image data. The recognition accuracy can be improved by accumulating feature amount data corresponding to a plurality of face images as dictionary data for one recognition candidate. In particular, recognition accuracy can be improved by accumulating feature amounts corresponding to a plurality of face images photographed under various photographing conditions that cause changes in the orientation, facial expression, lighting conditions, and other face images of the subject.

  The image data stored in the photo database 20 is, for example, photographic image data taken by a digital still camera, and mainly uses a person as a subject. The photographic image data may be acquired via a communication interface 12 such as a USB interface or a LAN interface, or via a memory card, CD-ROM or other recording medium (not shown) that can be loaded into the memory card slot 13. May be acquired.

FIG. 2 is a block diagram for explaining the configuration of the computer main body 1 in the execution state of the computer program. That is, the computer main body 1 substantially has a plurality of function processing units shown in FIG. 2 by executing the computer program.
The plurality of function processing units include a photo registration processing unit 30, a person image detection unit 31, a person image recognition processing unit 32 (person image recognition unit), a display signal generation processing unit 33 (display signal generation unit), and a correction input acceptance. Unit 34 (correction input receiving unit), recognition result correction processing unit 35 (recognition result correcting unit), confirmation input receiving unit 36 (confirmation input receiving unit), recognition result confirmation processing unit 37 (recognition result determination unit), search condition input It includes a reception unit 38, an image search processing unit 39 (search unit), a dictionary correction processing unit 40 (dictionary correction unit), and a recognition candidate exclusion processing unit 41 (recognition candidate exclusion unit).

The photo registration processing unit 30 executes processing for registering image data of a photo image in the photo database 20.
The person image detection unit 31 reads out image data stored in the photo database 20 and cuts out an image of a person's face part (image part to be processed by the person image recognition processing unit 32) in the image data. . More specifically, the person image detection unit 31 specifies a rectangular area including the face portion of the person in the image data as a face image area (cutout frame), and the origin coordinates of the face image area (for example, the upper left corner of the rectangular area) Data representing the corner vertex coordinates) and the cutting frame size (for example, the vertical and horizontal lengths of the rectangular area) are generated. These data are stored in the photograph database 20 in association with the image data.

  The person image recognition processing unit 32 performs a person recognition process on the face image area detected by the person image detection unit 31 and writes the processing result in the photo database 20. Specifically, the person image recognition processing unit 32 collates the image in the face image area with the dictionary data stored in the recognition dictionary storage unit 21, and determines a person candidate for the face image in the face image area. Generate candidate data to represent. More specifically, the person image recognition processing unit 32 relates to a plurality of recognition candidates (candidate persons) whose feature amount data is stored in the recognition dictionary storage unit 21 and is likely to be a person of a face image in the face image area. A likelihood calculating unit 43 (likelihood calculating unit) that calculates the likelihood to be expressed, and a priority setting unit 44 (priority setting unit) that determines the rank of recognition candidates based on the likelihood calculated by the likelihood calculating unit 43 ).

  The priority order setting unit 44 assigns ranks to recognition candidates in descending order of the likelihood, in principle, for face images in individual face image regions. However, when two or more face image areas are included in the same image data, the recognition candidate having the maximum likelihood in the plurality of face image areas may be the same person. In such a case, the likelihoods of the recognition candidates in the plurality of face image areas are compared, and the recognition candidate is set as the first candidate for the face image area with the higher likelihood, and other face images In the region, other recognition candidates are set as the first candidates.

The display signal generation processing unit 33 generates a display signal corresponding to an image to be displayed on the display device 3. Specifically, a display signal necessary for displaying image data and a recognition result of a person in the image data is generated.
The correction input receiving unit 34 receives an input for correcting the recognition result by the person image recognition processing unit 32. That is, when the user operates the input operation unit 2 and inputs an instruction for instructing correction of the recognition result, the instruction input is received by the correction input receiving unit 34. The correction input receiving unit 34 includes a selection candidate presenting unit 42 (selection candidate presenting means) for displaying recognition candidate options on the display device 3 in the form of a pull-down list box and presenting them to the user. The selection candidate presentation unit 42 gives a command for displaying the pull-down list box on the display device 3 to the display signal generation processing unit 33. Upon receiving this command, the display signal generation processing unit 33 generates a display signal for displaying the pull-down list box.

The recognition result correction processing unit 35 corrects the recognition result according to the instruction input received by the correction input receiving unit 34. More specifically, the recognition result correction processing unit 35 reads the recognition result data corresponding to the image data to be input as an instruction from the photo database 20 and executes a process for changing the recognition result.
The confirmation input receiving unit 36 receives a confirmation input for confirming and confirming a recognition result corresponding to each face image area stored in the photo database 20. That is, when the user operates the input operation unit 2 and inputs an instruction for instructing confirmation of the recognition result, the instruction input is received by the confirmed input receiving unit 36.

A confirmation flag as confirmation completion data indicating whether the recognition result is confirmed by the user (confirmed) or not confirmed (unconfirmed) for each image data is stored in the photo database 20. It has become so. Immediately after the recognition processing by the person image recognition processing unit 32, the confirmation flag is set to a value representing “unconfirmed”.
The recognition result confirmation processing unit 37 performs a process of changing the confirmation flag for each individual face image area recorded in the photo database 20 to “confirmed”. The recognition result confirmation processing unit 37 rewrites the confirmation flag to “confirmed” in response to the confirmation input accepted by the confirmation input acceptance unit 36. Further, even when the recognition result correction processing unit 35 corrects the recognition result, the recognition result confirmation processing unit 37 rewrites the confirmation flag corresponding to the face image area to be corrected to “confirmation”. The face image area in which the value of the confirmation flag is “confirmed” is excluded from the processing target of the person image recognition processing unit 32.

Furthermore, the recognition result confirmation processing unit 37 includes a recognition result automatic confirmation processing unit 45 (recognition result automatic confirmation means). The recognition result automatic confirmation processing unit 45 automatically changes the confirmation flag corresponding to the face image area for which the recognition result is undetermined to “confirmed” when a certain condition is satisfied.
More specifically, a plurality of pieces of image data stored in the photo database 20 are ordered according to a predetermined standard. For example, they may be sorted in the order of the shooting date of the photos, may be sorted in the order of registration in the photo database 20, may be sorted in the order of the last update date and time of the image data, It may be sorted by data name (file name). In such a case, when the user performs a confirmation / or correction operation with respect to certain image data, the recognition result automatic confirmation processing unit 45 determines that the recognition result is unconfirmed that exists in the image data of the rank before the image data. The confirmation flag is changed to “confirmed” for all face image regions.

The search condition input accepting unit 38 accepts a search condition specified by the user operating the input operation unit 2. As search conditions, identification information of recognition candidates recognized by the person image recognition processing unit 32 can be set. This identification information is information registered as a recognition result in the photo database 20, and a typical example is a name of a recognition candidate (last name, first name, common name, etc.).
The image search processing unit 39 searches for image data according to the search condition received by the search condition input receiving unit 38. For example, when the user designates one of the names of recognition candidates as a search condition, the image search processing unit 39 extracts image data including a face image area having the recognition candidate name as a recognition result.

  The dictionary correction processing unit 40 corrects the dictionary data stored in the recognition dictionary storage unit 21 in response to the confirmation process by the recognition result confirmation processing unit 37. More specifically, when the recognition result is confirmed by the recognition result confirmation processing unit 37, the correspondence between the confirmed recognition result and the face image data (image data in the face image area) is reflected in the dictionary data. It is done. That is, feature amount data corresponding to the face image data is added to the dictionary data for recognition candidates corresponding to the recognition result. Thus, as the recognition process is performed, the accuracy of the dictionary data is improved.

In particular, in this embodiment, when the recognition result is corrected by the recognition result correction processing unit 35, the recognition result is fixed. Therefore, the correction of the recognition result can be reflected in the dictionary data. As a result, the accuracy of the dictionary data is efficiently increased.
FIG. 3 is an illustrative view for explaining the structure of data stored in the photo database 20. The photo database 20 stores photo data i (i = 1, 2,..., N, where N is a natural number and represents the total number of photo data) corresponding to a plurality of photos. . In this case, each photograph data i includes image data corresponding to the photograph image and additional data associated with the image data. The additional data includes number data representing the number M (M is a natural number) of face image areas detected in the image data, and individual area data j (j = 1, 2,..., Each face image area). M). The individual area data j includes coordinates of the face image area, a cutout frame size, recognition result data, recognition candidate data, and a confirmation flag. The recognition result data is identification information for identifying the person of the recognition result, and may be, for example, the name of the person. Further, the recognition candidate data includes identification information for specifying the recognition candidate and a likelihood representing the likelihood that the person in the face image region is the recognition candidate with respect to at least one (usually a plurality of) recognition candidates. Degree data.

In a state immediately after the recognition processing by the person image recognition processing unit 32, the first recognition candidate data and the recognition result data match. The correction process by the recognition result correction processing unit 35 is a process for rewriting the recognition result data.
FIG. 4 is an illustrative view showing a display example on the display device 3. In this example, photographic images 51, 52, 53 and 54 are arranged and displayed in the display window 50. These photographic images 51 to 54 are ranked in this order. FIG. 4 shows a situation where the user performs a correction operation on the recognition result for the third photographic image 53.

  A scroll bar 55 is arranged on the right side of the display window 50. By operating the scroll bar 55 with a pointing device or by performing a predetermined scroll operation from the keyboard, the photographic images ranked before the photographic images 51 to 54, or the photographic images ranked after these. Can be displayed on the display window 50. An end button 56 for ending the program is arranged at the upper right corner of the display window 50.

  In the display window 50 in the state shown in FIG. 4, the first photographic image 51 includes a whole body image of one person A. A rectangular face image area 61 including the face portion of the person A is detected by the person image detecting unit 31. In the vicinity of the face image area 61, a confirmed recognition result 62 is displayed. In this example, the recognition result 62 is a character string “Mr. A” representing the name of the person A. The feature amount data corresponding to the face image data in the face image area 61 is written into the recognition dictionary storage unit 21 as dictionary data relating to the recognition candidate “Mr. A” corresponding to the confirmed recognition result 62.

  The second photographic image 52 includes full body images of two persons A and B. The person image detection unit 31 detects two face image areas 63 and 64 each including the face image portions of the persons A and B. In the vicinity of these face image areas 63 and 64, confirmed recognition results 65 and 66 are displayed, respectively. The recognition results 65 and 66 are character strings “Mr. A” and “Mr. B”, respectively. The feature amount data corresponding to the face image data in the face image areas 63 and 64 is recognized as dictionary data associated with the recognition candidates “Mr. A” and “Mr. B” corresponding to the recognized recognition results 65 and 66, respectively. It is written in the dictionary storage unit 21.

  The third photographic image 53 includes a whole body image of one person 67. The person image detection unit 31 extracts a rectangular face image area 68 including the face portion of the person 67. A recognition candidate 69 is displayed in the vicinity of the face image area 68. The recognition candidates 69 are displayed in the form of a pull-down list box 70. That is, in the initial state, the pull-down list box 70 is a state in which only one candidate with the maximum likelihood is displayed. From this state, a correction candidate can be displayed by operating (clicking) the pull-down button 71 at the right end with a pointing device.

  In this example, the maximum likelihood candidate is “Mr. A”, and the second most likely candidate is “Mr. B”. The operation for selecting the maximum likelihood candidate “Mr. A” is a confirmation operation for confirming the automatically recognized recognition result. In addition, an operation for selecting a candidate other than the maximum likelihood is a correction operation for correcting the automatically recognized recognition result. When a confirmation operation or a correction operation is performed, the recognition result becomes a confirmation state. As a result, the feature amount data corresponding to the face image data in the face image area 68 is stored in the recognition dictionary storage unit 21 as dictionary data relating to recognition candidates corresponding to the determined recognition result.

In the pull-down list box 70, in addition to the recognition candidates, options of “new registration” (new registration means) and “no registration” (registration means without additional information) are prepared.
“New registration” is an option for starting an operation of registering a person who is not registered as a recognition candidate in the dictionary data of the recognition dictionary storage unit 21. When “new registration” is selected, the “new registration” field in the pull-down list box 70 becomes an editable text input area. By inputting identification information (such as a name) of recognition candidates from the input operation unit 2 in this text input area, the recognition result of the face image area 68 is corrected and confirmed. At the same time, the feature amount data corresponding to the face image data in the face image area 68 is stored in the recognition dictionary storage unit 21 as dictionary data associated with the newly input recognition candidate (for example, “Mr. C”). The

“No registration” is selected, for example, as a recognition result of an unacquainted person. More generally, it is appropriate that the recognition result is “no registration” (no additional information) for a face image region of a person with a very low probability of being searched later.
The fourth photographic image 54 includes full body images of four persons 75, 76, 77, 78. Among these persons 75 to 78, face image areas 80, 81, and 83 are detected for persons 75, 76, and 78, and recognition candidates corresponding to these are displayed in pull-down list boxes 85, 86, and 88. It is displayed in the format. For the person 77, the face image area has not been detected. However, by pointing the person 77 using a pointing device, additional data for the person 77 can be registered afterwards.

  Assume that any one of the first to third photographic images 51 to 53 includes a face image area in which the recognition result is undetermined. In this case, when the user performs a correction operation for correcting the recognition result or a confirmation operation for confirming the recognition result on the face image area in the fourth photographic image 54, the recognition result automatic confirmation processing unit 45 The recognition result of the unconfirmed state in the first to third photographic images 51 to 53 is confirmed by the operation.

  However, after the recognition result is confirmed, the confirmed recognition result can be corrected. For example, when a predetermined operation (click) is performed by pointing the recognition result 62 displayed near the face image area 61 of the first photographic image 51, a list of recognition candidates is displayed in the form of a pull-down list box. . Appropriate recognition candidates (including “new registration” and “no registration”) can be selected from this list.

As described above, when a correction or confirmation operation or a new registration operation is executed on the recognition result, the dictionary data in the recognition dictionary storage unit is updated. Therefore, using the updated dictionary data, the person image recognition process is performed again on the photographic images in the rank after the photographic image on which the operation for correction or confirmation or new registration is executed.
FIG. 5 is a diagram for explaining more detailed processing when the recognition result is corrected. In FIG. 5, the photographic image 90 includes full-length images of two persons 91 and 92, and face image areas 93 and 94 of these persons 91 and 92 are detected. For the person 91, the recognition result 95 has been confirmed as “Mr. B”. In this case, “Mr. B” is excluded from the list in the pull-down list box 96 indicating recognition candidates for the face image area 94 of the person 92.

This is a process based on the fact that, in principle, a face image of the same person never appears twice in the same photograph, and is based on the action of the recognition candidate exclusion processing unit 41 (see FIG. 2). As a result, it is possible to properly perform person recognition on each face image area in the photographic image, and to smoothly perform correction of the recognition result.
FIG. 6 is a diagram for explaining the details of the processing related to the display order of recognition candidates. In FIG. 6, the photographic image 100 includes full body images of three persons 101, 102, and 103, and face image areas 104, 105, and 106 of these persons 101, 102, and 103 are detected. Yes. An example of the likelihood calculated by the human image recognition process based on the images included in these face image areas 104, 105, and 106 is shown in Table 1 below.

Immediately after the person image recognition processing, the first-ranked recognition candidates (that is, recognition candidates with the maximum likelihood) are displayed in the pull-down list boxes 107, 108, and 109 for the face image regions 104, 105, and 106, respectively. The If the first recognition candidate overlaps in a plurality of face image regions, the higher likelihood is given priority.
For example, in the example of FIG. 6, processing when the recognition result of the face image area 104 is changed from “Mr. A” to “Mr. B” will be described. In the initial recognition result, the character string “Mr. B” is displayed in the pull-down list box 108 as a recognition candidate for the face image area 105. However, the recognition candidate “Mr. B” in the face image area 105 inevitably becomes an error because the recognition result of the face image area 104 is corrected and determined to be “Mr. B”. This is because it is impossible in principle for the same person to appear twice in the same photographic image.

  Therefore, the priority order setting unit 44 corrects the recognition candidates for the face image areas 105 and 106. In this case, referring to Table 1, the second recognition candidate in the face image area 105 is “Mr. C”, which overlaps with the first recognition candidate in the face image area 106. Therefore, the priority order setting unit 44 compares the likelihood corresponding to the recognition candidate “Mr. C” in the face image areas 105 and 106. Referring to Table 1, the likelihood of the recognition candidate “Mr. C” in the face image area 105 is larger. Therefore, the priority order setting unit 44 changes the recognition candidate of the face image area 105 to “Mr. C”. At the same time, the priority order setting unit 44 excludes the confirmed “Mr. B” as the recognition candidate for the face image area 106 and the “Mr. C” as the recognition candidate for the face image area 105. , Change to the most likely recognition candidate “Mr. A”. Of course, the confirmed candidate “Mr. B” is excluded from the recognition candidate choices in the face image areas 105 and 106 by the action of the recognition candidate exclusion processing unit 41.

  FIG. 7 is a flowchart for explaining the flow of processing. First, the user registers the image data of a photographic image in the photographic database 20 by operating the input operation unit 2 (step S1). In response to the operation of the input operation unit 2, the photo registration processing unit 30 stores in advance in a storage area of the hard disk drive 8, a CD-ROM, a memory card or other external recording medium, or a predetermined location on the network, for example. The registered image data is registered in the photo database 20. More specifically, photo data including image data (see FIG. 3) is created for the number of image data to be registered and registered in the photo database 20.

  Next, the person image detection unit 31 performs a face image area detection process on the image data registered in the photograph database 20 (step S2). At this time, the target of the face image area detection process is image data that has not undergone the face image area detection process. In this way, face image areas are extracted from all image data for which face image area detection processing has not been completed, and data representing the coordinates of each face image area and the cutout frame size is registered in the photo database 20.

Next, the person image recognition process by the person image recognition processing unit 32 is executed on all the face image areas detected by the face image area detection process in step S2 (step S3). Thereby, recognition candidate data for each face image area is registered in the photo database 20. The recognition result is displayed on the display device 3 together with the photographic image (step S4).
After the person image recognition process (step S3) is completed, the user operates (clicks) any position on the display screen of the display device 3 with the pointing device by operating the input operation unit 2 (step S5). The CPU 5 determines whether to end the program (step S6). That is, it is determined whether or not to end the program depending on whether or not the end button 56 on the screen (see FIG. 4 may be an end button separately displayed in the window) is operated.

When the program is to be ended, the display signal generation processing unit 33 displays a signal for causing the display device 3 to display a selection screen for allowing the user to select “confirm” or “not confirm” the recognition result. Generate (step S7). This selection screen may be, for example, a window including a “confirm” button and a “not confirm” button.
When the user performs a selection operation for confirming / non-confirming the recognition result from the input operation unit 2 (step S8), processing corresponding to the selection operation is executed by the CPU 5 (step S9). That is, when the “confirm” button is operated (YES in step S9), this operation is accepted by the confirmation input receiving unit 36, and the recognition result confirmation processing unit 37 confirms all of the unconfirmed recognition results ( Step S10). That is, all the “unconfirmed” confirmation flags are changed to “confirmed”. Thereby, the process ends. When the “don't confirm” button is operated (NO in step S9), the process of step S10 is bypassed and the process ends. In this case, the confirmation flag of the face image area whose recognition result is undetermined remains “undetermined”.

  On the other hand, in step S5, when the operation position (click position) is a place other than the end button 56 (however, in the window 50), it is determined whether or not the position is in the same photographic image as the previous operation position (click position). (Step S11). If the operation position is not within the same photo image, the recognition result automatic confirmation processing unit 45 sets the confirmation flag to “Confirm” for all face image regions in all the photo images ranked before the photo image. Is performed (step S12). If the operation position is not within the same photograph image, the recognition result automatic determination process in step S12 is omitted.

  Next, the CPU 5 determines whether or not the operation position (click position) is a recognition candidate display position (step S13). When the operation position is the recognition result display position (the above-described pull-down list box), the CPU 5 executes a “candidate selection button click process” (step S14). If the operation position is not the recognition result display position, the CPU 5 executes “undetected position click process” when a position other than the position where the face image area is detected is clicked (step S15).

  FIG. 8 is a flowchart for explaining details of the “processing when the candidate selection button is clicked” (step S14 in FIG. 7). As described above, the recognition result is displayed in the form of a pull-down list box by the function of the selection candidate presentation unit 42 (see FIG. 4 and the like). When this pull-down list box is clicked, a pull-down list showing options including recognition candidates is expanded and displayed (step S21). As described above, the recognition candidates are basically displayed in the order of likelihood. In addition to the recognition candidates, “new registration” and “no registration” are displayed as options.

When the user performs an operation (click) for selecting one of the options by operating the input operation unit 2 (step S22), this operation is received by the correction input receiving unit 34 or the confirmed input receiving unit 36. The result is transferred to the recognition result correction processing unit 35 or the recognition result confirmation processing unit 37.
When any recognition candidate is selected (YES in step S23), the recognition result correction processing unit 35 or the recognition result confirmation processing unit 37 displays the recognition result corresponding to the face image region as the selected recognition candidate. Perform the process to confirm. More specifically, when the first recognition candidate is selected (YES in step S24), this selection input is accepted by the confirmation input acceptance unit 36, and the recognition result confirmation processing unit 37 recognizes the recognition result of the face image region. In order to confirm (equal to the first recognition candidate), the value of the confirmation flag is rewritten to “confirmed” (step S26). On the other hand, when the second or lower recognition candidate is selected (NO in step S24), this selection input is received by the correction input receiving unit 34, and the recognition result correction processing unit 35 selects the selected recognition candidate. The recognition result of the face image area is written in the photo database 20 (step S25). Further, upon receiving the notification of the correction processing by the recognition result correction processing unit 35, the recognition result determination processing unit 37 rewrites the value of the determination flag of the face image area to “confirmation” (step S26).

  On the other hand, when “new registration” is selected (NO in step S23 and YES in step S27), the display signal generation processing unit 33 allows the user to display the name of the person in the face image and other identification information. A text input area for inputting is displayed on the display device 3 (step S28). When the user operates the input operation unit 2 to input a name and other identification information to the text input area (step S29), the recognition result correction processing unit 35 uses the input identification information as the face. The image area recognition result is registered in the photo database 20 (step S30). Receiving this registration notification, the recognition result confirmation processing unit 37 rewrites the value of the confirmation flag corresponding to the face image area to “confirm” (step S26).

Further, when “no registration” is selected (NO in step S23 and NO in step S27), the recognition result correction processing unit 35 sets the recognition result of the face image area as “no registration” in the photo database 20. Writing (step S31), the recognition result confirmation processing unit 37 rewrites the value of the confirmation flag to “confirm” (step S32).
In step S26, after the recognition result of the photo database 20 is confirmed, the dictionary correction processing unit 40 performs correction processing of the dictionary data in the recognition dictionary storage unit 21 (step S33). That is, the dictionary correction processing unit 40 adds the feature amount data of the image in the face image area to the recognition dictionary storage unit 21 as dictionary data related to the recognition candidate corresponding to the confirmed recognition result.

Thereafter, the CPU 5 executes the “duplicate processing within the same photo” described later (step S34), and further uses the corrected dictionary data to all uncertain face image areas after the photo image being processed. On the other hand, a human image recognition process is executed (step S35).
FIG. 9 is a flowchart for explaining details of the “processing when clicking an undetected position” (step S15 in FIG. 7). This process is a process for adding additional data for search to an arbitrary position where the face image area is not detected. When the user performs an operation (click) to select a position in which no face image is detected in the photographic image, the selection candidate presenting unit 42 has a plurality of registration positions registered in the dictionary data at the operation position (click position). A command for displaying the pull-down list box including the recognition candidates “new registration” and “no registration” in the expanded form is given to the display signal generation processing unit 33 (step S40). In this state, the user operates the input operation unit 2 to select one of the options (step S41). If the operated option is any recognition candidate (YES in step S42), the recognition result correction processing unit 35 recognizes the selected recognition candidate including the coordinates of the click position (step S5 in FIG. 7). Individual area data as a result (additional data for search) is generated and registered in the photo database 20 (step S43). Receiving this registration notification, the recognition result confirmation processing unit 37 sets the value of the confirmation flag of the individual area data to “confirm” (step S44).

  On the other hand, when the “new registration” option is selected (NO in step S42 and YES in step S45), the correction input receiving unit 34 inputs a text for inputting identification information of a recognition candidate to be newly registered. The area is displayed on the display screen of the display device 3 (step S46). When the user inputs identification information such as a name by operating the input operation unit 2 in this text input area (step S47), the recognition result correction processing unit 35 includes the coordinates of the click position in step S5 of FIG. Individual area data having the identification information as a recognition result is generated and registered in the photograph database 20 (step S48). Receiving this registration notification, the recognition result confirmation processing unit 37 changes the value of the confirmation flag of the individual area data to “confirm” (step S44).

  When the “no registration” option is selected (NO in step S42 and NO in step S45), the recognition result correction processing unit 35 includes the coordinates of the click position (step S5 in FIG. 7), and the recognition result. The individual area data with “not registered” is generated and registered in the photo database 20 (step S49). Further, the recognition result confirmation processing unit 37 sets the value of the confirmation flag of the individual area data to “confirm” (step S44).

  FIG. 10 is a flowchart for explaining the details of the human image recognition process (step S3 in FIG. 7 and step S35 in FIG. 8). The person image recognition processing unit 32 sequentially processes the photographic images including the face image region whose confirmation flag value is “unconfirmed”, and the image in the face image region is stored in the recognition dictionary storage unit 21 and the dictionary data. By collating, the person image is recognized (step S51). Thereby, the likelihood that the person in the face image area is the recognition candidate is calculated for all the recognition candidates stored in the dictionary data. The calculated likelihood for each recognition candidate is registered as recognition candidate data in the individual area data corresponding to the face image area of the photo database 20.

Next, the “duplicate processing in the same picture” described later (step S52) is performed.
When the person image recognition processing unit 32 performs the same processing for all the face image areas in one photographic image, it checks whether or not the photographic image is the last photographic image (step S53). If not, the same process is repeated for the next photographic image (step S54). Thus, when the process proceeds to the last photo and the process ends (YES in step S53), the display signal generation processing unit 33 generates a display signal for displaying the recognition result on the display device 3 (step S55). . Thereby, the person image recognition process is completed.

  FIG. 11 is a flowchart for explaining details of the “duplicate processing in the same picture” (step S34 in FIG. 8 and step S52 in FIG. 10). When there is a recognition result of the same person in the same photograph (step S60), the priority order setting unit 44 executes a process for correcting this. That is, for example, when the recognition result overlaps for a pair of face image areas that follow each other in the order of face image area detection (YES in step S60), the priority order setting unit 44 determines one of them. It is determined whether or not (step S61). If one is confirmed, the recognition candidate with the highest likelihood among the recognition candidates not selected as the recognition result of the face image region confirmed in the same photograph for the face image region that has not been confirmed. The recognition result is changed to (Step S62).

  On the other hand, when the recognition results overlap (YES in step S60), if both of them are uncertain, the likelihoods of those recognition candidates are compared, and the recognition result with the higher likelihood is given priority, For the recognition result with the lower likelihood, the recognition candidate with the next highest likelihood is corrected as the recognition result (step S62). If the recognition results overlap and their likelihoods are equal, for example, priority is given to the recognition result of the face image detection process performed first (the person recognition process performed first). In other words, the recognition result for the face image area processed later may be corrected.

  As described above, according to this embodiment, a human image recognition process is performed by a computer, and the user corrects the recognition result by operating the input operation unit 2. Therefore, the person image recognition process does not need to be complete, and accurate additional information (person information) can be given to all the photographic images by a correction input by the user. Further, by utilizing the human image recognition process by the computer, the input burden on the user is reduced, and accurate additional information can be added to the photographic image with less effort. By setting the additional information correctly given as a search condition in this way, an accurate search process can be executed.

  Further, in this embodiment, the dictionary data in the recognition dictionary storage unit 21 is corrected by confirming or correcting the recognition result by the user. As a result, the more photographic images are accumulated, the higher the accuracy of the human image recognition process. As a result, the correction work of the human image recognition processing result (recognition result) decreases as the user uses it, and the user's labor can be reduced.

Further, since the recognition result is automatically corrected based on the fact that the same person's face image does not appear in the same photo image, a reasonable recognition result can be provided. Thereby, the user's correction burden can be further reduced.
Furthermore, when the confirmation or correction process is executed on any of the photographic image data, the recognition result is automatically determined for the face image areas in the photographic image data of the previous rank. Therefore, the user does not need to perform an operation for determining the recognition result for all the face image areas, and only needs to perform a correction operation for an image that requires correction of the recognition result. Thereby, the operation burden on the user can be further reduced.

The recognition result is determined for the purpose of specifying the processing target by the human image recognition processing unit 32. That is, the person image recognition processing unit 32 executes the person image recognition process only for the face image area whose recognition result is undetermined. Thereby, it is possible to prevent the correct recognition result from being changed to an incorrect recognition result while the person image recognition process is repeatedly executed.
As mentioned above, although one Embodiment of this invention was described, this invention can also be implemented with another form.

  For example, in the above-described embodiment, the names of recognition candidates are registered as recognition results. However, data specifying typical face image data is registered in the photo database 20 as recognition result data. May be. Specifically, index image data representing a representative face image of a recognition candidate may be registered in the photo database 20 as indicated by a virtual line in FIG. That is, the photo database 20 is used as index image data storage means, and an index image registration receiving unit 131 and an index image registration processing unit 132 are provided. When the user performs a predetermined input operation for registering an index image on the input operation unit 2, this input is received by the index image registration receiving unit 131, Index image data is registered in the photo database 20.

  Thereby, for example, by performing a predetermined input operation on the screen as shown in FIG. 4, the face image data in the face image area 61 is registered in the photo database 20 as index image data of the recognition candidate “Mr. A”. can do. Further, the data in the face image area when newly registered may be registered in the photo database 20 as index image data.

  As the recognition result, identification data for specifying the index image data is generated by the person image recognition processing unit 32 and registered in the individual area data. In this case, the display signal generation processing unit 33 reads the index image data from the photo database 20 and generates a display signal for displaying the index image data specified by the identification data in the vicinity of each face image region. It is preferable. The selection candidate list may also be displayed using the index image data, and the search condition may be set using the index image data.

With such a configuration, the recognition result can be grasped and corrected by an intuitive operation, and a necessary photographic image can be searched. In addition, it is possible to manage photographic images by operating only a pointing device without using a skillful input device such as a keyboard.
Furthermore, in the above-described embodiment, the face image area is detected and the person is recognized by the face image recognition process. However, the person recognition process is not limited to this. For example, recognition based on the color of a person's clothes is also possible. In this case, the process performed by the person image detection unit 31 is a process of cutting out an image of a person's clothes.

In the above-described embodiment, an example is shown in which a full body image of a person is shown in a photographic image, but it is needless to say that a full body image of the person is not necessarily shown in the photographic image.
Furthermore, in the above-described embodiment, an example in which an image management program is executed on a personal computer has been described. However, the present invention can be applied to an application program on the web (document system used on a network represented by the Internet and an intranet). The image management apparatus may be configured on a server which is a computer placed on a network. In this case, the user operates a client device, which is another computer connected to the network, to perform operations such as registration of photographic images, correction / confirmation of recognition results, and registration of new recognition candidates via the network. It can be carried out.

Of course, the image management apparatus of the present invention does not need to be configured using a general-purpose computer, and can also be configured as a dedicated apparatus capable of stocking and browsing photographic images. Such a dedicated device can be installed in a photographic image printing service store and incorporated into a photographic image print ordering system, for example.
In addition, various design changes can be made within the scope of matters described in the claims.

It is a block diagram for explaining a basic configuration of a computer operable as an image management apparatus according to an embodiment of the present invention. It is a block diagram for demonstrating the functional structure of the computer main body in the execution state of the said computer program. It is an illustration figure for demonstrating the structure of the data stored in a photograph database. It is an illustration figure which shows the example of a display in a display apparatus. It is a figure for demonstrating the more detailed process at the time of recognition result correction. It is a figure for demonstrating the detail of the process regarding the display order of a recognition candidate. It is a flowchart for demonstrating the flow of a process. It is a flowchart for demonstrating the detail of "a candidate selection button click process" (step S14 of FIG. 7). It is a flowchart for demonstrating the detail of "the process at the time of an undetected position click" (step S15 of FIG. 7). It is a flowchart for demonstrating the detail (step S3 of FIG. 7, and step S35 of FIG. 8) of a person image recognition process. 13 is a flowchart for explaining details of “duplicate processing in the same photograph” (step S34 in FIG. 8 and step S52 in FIG. 10).

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Computer main body 2 Input operation part 3 Display apparatus 5 CPU
6 ROM
7 RAM
8 Hard disk drive 9 CD-ROM drive 10 Display controller 11 Input controller 12 Communication interface 13 Memory card slot 14 Bus 15 CD-ROM
DESCRIPTION OF SYMBOLS 20 Photo database 21 Recognition dictionary memory | storage part 30 Photo registration process part 31 Person image detection part 32 Person image recognition process part 33 Display signal generation process part 34 Correction input reception part 35 Recognition result correction process part 36 Confirmation input reception part 37 Recognition result confirmation Processing unit 38 Search condition input receiving unit 39 Image search processing unit 40 Dictionary correction processing unit 41 Recognition candidate exclusion processing unit 42 Selection candidate presentation unit 43 Likelihood calculation unit 44 Priority order setting unit 45 Recognition result automatic decision processing unit 50 Display window 51 -54 Photo image 55 Scroll bar 56 End button 61 Face image area 62 Recognition result 63, 64 Face image area 65, 66 Recognition result 67 Person 68 Face image area 69 Recognition candidate 70 Pull-down list box 71 Pull-down button 75-78 Person 80, 81, 83 Face image area 85, 86, 8 Pull-down list box 90 Photo image 91, 92 Person 93, 94 Face image area 95 Recognition result 96 Pull-down list box 100 Photo image 101-103 Person 104-106 Face image area 107-109 Pull-down list box 131 Index image registration processing unit 132 Index Image registration processing part A, B

Claims (11)

Image data storage means for storing a plurality of image data including a person image;
Person image recognition means for recognizing person images in individual image data stored in the image data storage means;
A recognition result storage means for storing the recognition result by the person image recognition means in association with each image data;
Correction input receiving means for receiving a correction input for correcting a recognition result by the person image recognition means;
An image management apparatus comprising: a recognition result correcting unit that corrects a recognition result stored in the recognition result storage unit according to a correction input received by the correction input receiving unit.   2. The display signal generating means for generating a display signal for displaying a recognition result by the human image recognition means on a display device together with the image data stored in the image data storage means. Image management device. Dictionary storage means for storing dictionary data for recognition of person images;
A dictionary correction unit for correcting dictionary data in the dictionary storage unit based on the recognition result corrected by the recognition result correction unit;
3. The image management apparatus according to claim 1, wherein the person image recognition means recognizes a person image in the image data based on dictionary data stored in the dictionary storage means. A recognition result confirmation unit for confirming the recognition result stored in the recognition result storage unit;
4. The image management apparatus according to claim 3, wherein the dictionary correction means includes means for correcting dictionary data in the dictionary storage means based on the determined recognition result.   5. The image management apparatus according to claim 1, further comprising search means for searching for image data based on a recognition result stored in the recognition result storage means. The person image recognition means
Likelihood calculating means for calculating the likelihood that the person image is an image of a recognition candidate for each of the plurality of recognition candidates;
Priority for determining the priority of recognition candidates according to the likelihood obtained by the likelihood calculating means when the maximum likelihood recognition candidate overlaps between two or more person images in image data including a plurality of person images 6. The image management apparatus according to claim 1, further comprising a rank setting unit. A recognition result confirmation unit for confirming the recognition result stored in the recognition result storage unit;
The correction input receiving means includes a selection candidate presenting means for presenting a plurality of recognition candidate options to the user,
When the recognition result of any person image is determined by the recognition result determination means in the image data including a plurality of person images, the selection candidate presenting means for the person image whose recognition result is not determined in the image data. The image management apparatus according to claim 1, further comprising: a recognition candidate exclusion unit that excludes the confirmed recognition result from the recognition candidate options presented by the step.   8. The image according to claim 1, wherein the person image recognition means generates index image specifying data for specifying index image data corresponding to the recognition result as the recognition result data. Management device. A recognition result confirmation unit for confirming the recognition result stored in the recognition result storage unit;
The image data stored in the image data storage means are ordered,
The person image recognition means performs person image recognition processing on a plurality of pieces of image data stored in the image data storage means according to the ordering,
9. The automatic recognition result confirming unit for automatically confirming a recognition result related to image data of an order prior to image data corresponding to the recognition result determined by the recognition result determining unit. The image management apparatus described in 1. A computer program for operating a computer as an image management device for managing a plurality of image data including a person image,
The computer includes image data storage means for storing a plurality of image data including a person image, and a recognition result storage means for storing a recognition result of the person image,
The computer program stores the computer,
Person image recognition means for recognizing a person image in individual image data stored in the image data storage means;
Means for writing the recognition result by the person image recognition means in association with each image data into the recognition result storage means;
A correction input receiving means for receiving a correction input for correcting the recognition result by the person image recognition means, and a recognition result stored in the recognition result storage means in response to the correction input received by the correction input receiving means. A computer program for image management, which functions as a recognition result correcting means for correcting the image.   A computer-readable recording medium on which the computer program according to claim 10 is recorded.