JP2003271612A - Image database creation system and retrieval system - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an image database creation system and retrieval system capable of performing a precise and efficient image retrieval. <P>SOLUTION: A concept index indexA related to image data Di to be registered and created by a concept index creation means 16 is standardized to image data D accumulated in an image database 200 by a concrete index creation means 16 to create a concrete index indexXb. The concrete index indexXb is standardized to a database other than the image database by a registered index creation means 17 to create a concrete index indexZb to be registered in the image database. <P>COPYRIGHT: (C)2003,JPO

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention creates a concept index of image data such as still images, accumulated content data, and metadata published on the Internet. Create a concrete index for related matters,
The present invention relates to an image database creation system that stores the created conceptual index and concrete index in association with the above-mentioned image data, content data, and metadata.

Further, according to the present invention, desired image data, content data, and metadata published on the Internet are extracted from the image database thus created on the basis of the above-mentioned concrete index or concept index. The present invention relates to an image database search system for searching.

[0003]

2. Description of the Related Art As a conventional image database search system, color information, appearance information such as shape, size, orientation, etc. of an image to be searched from a huge amount of accumulated image data, on-screen Primary index information that can be directly derived from the image to be searched, such as position information of where the target exists (hereinafter,
There was an image database search system that searches for images to be searched using (concept index information). Then, according to this image database search system, a database user (searcher) can perform a search based on image data at the time of search. Therefore, even if the database user does not have special knowledge about the image to be searched, as long as he / she has the image data as the search key, the user can perform the search.

Further, as another conventional image database search system, a huge amount of accumulated image data is used to acquire information such as the date and time of the photograph and the photographer, and an index given to the image itself by the database registrant. information,
There was an image database search system that searches for images to be searched using the above. According to this image database search system, a database user (searcher) can perform a search based on secondary index information (hereinafter referred to as concrete index information) attached to each image. Because of this, high-speed search can be performed.

[0005]

However, in the case of the former image database search system, when the database searcher does not have the image data to be the search key, the image data to be the basis of the search is obtained. The feature index information such as color information, appearance information such as shape, size, orientation, etc. of the image as a search material, and position information of the search target position on the screen, etc. However, there is a problem that the image to be searched cannot be searched because the image cannot be acquired. In the former image database search system of the related art, the image data registered in the image database is HTML (hypertext marku) published on the Internet.
If it was created in p language), the database registrant needed to download and register only the target image data.

The image data registered in the image database is SGML (standard generalized markup l).
In the case of anguage) and XML (extensible markup language), it takes a long time to search because all registered SGML and XML data are searched at the time of searching. Furthermore, in a system that creates a concept index from pre-stored image data, the color, direction (appearance), and
If you search by location, etc., the number of target items will increase,
It took a long time to search.

On the other hand, in the case of the latter conventional image database search system in which the database registrant or the like searches based on the specific index information given to the image itself, the specific index is used due to the difference in the concept of the index itself. Information cannot be normalized (standardized), and subtle deviations in the index information may hit image data other than the search target, or search omission may occur. SUMMARY OF THE INVENTION It is an object of the present invention to solve the above-mentioned problems of the conventional image database search system, and to provide an image database creation system and a search system capable of performing accurate and efficient image search.

[0008]

The present invention has been made to solve the above problems. First, the image database creating system of the present invention is based on the concept of the image data to be registered in the image database. Based on the concept index creating means for creating an index, and the concept index created by the concept index creating means, the image data registered in the image database is searched, and the specific index of the searched image data is found. A database other than the image database is extracted on the basis of a specific index creating unit that determines the credit rating of the extracted specific index and the specific index with the credit rating determined by the specific index creating unit. Search for the specific information that the searched data has A registration index for extracting a security index for each of the extracted specific indexes and setting a specific index to be stored in the image database in association with the image data to be registered, based on the determination result. And a creating unit.

Further, the image database creating system of the present invention is characterized in that the databases other than the image database are composed of a dictionary database connected via a network or metadata on the Internet. Further, the credibility determination by the specific index creating unit and the registration index creating unit is performed based on the degree of overlap of the specific indexes of the searched data among the searched data. Further, the registration index creating means outputs the specific index set corresponding to the image data to be registered in order to register the specific index in the image database in order to combine and output the credibility of the specific index set. It is characterized by doing. As a result, according to the image database creating system of the present invention, it is possible to create an image database that enables accurate and efficient image retrieval.

Next, the image database search system of the present invention comprises a concept index creating means for creating a concept index from the characteristics of image data to be registered in the image database, and a concept index created by the concept index creating means. A specific index creating unit that searches the image data registered in the image database, extracts a specific index included in the searched image data, and determines the credibility of the extracted specific index; A database other than the image database is searched on the basis of the specific index with the credibility determined by the specific index creating means, the specific index included in the searched data is extracted, and the extracted specific index Judging the credibility of each index, Based on the determination result, a keyword from an image database created by an image database creating system including a registration index creating unit that sets a specific index to be stored in the image database in association with the image data to be registered A feature is that a desired image is searched by any one of the three search categories of input, image data input, and URL input. As a result, according to the image database creation system of the present invention, a desired image can be accurately and efficiently searched by any one of the three search categories of keyword input, image data input, and URL input. You can do it.

[0011]

DETAILED DESCRIPTION OF THE INVENTION Referring to the accompanying drawings,
An image database creation / search system according to a preferred embodiment of the present invention will be described in detail. FIG. 1 is a block diagram showing the configuration of an image database creation / search system according to an embodiment of the present invention. In FIG.
Image database creation / search system 1 according to the present embodiment
Is a display unit 11, an input unit 12, an image data reading unit 13, a display control unit 14, a conceptual index creating unit 15, a concrete index creating unit 16, a registration index creating unit 17, and a category classification creating. Means 18
And image data search means 19.

The display unit 11 is composed of a display or the like, and displays the search results by the image database 200, which will be described later, etc. The input unit 12 is composed of a keyboard having kana keys and function keys, a pointing device such as a mouse, and the like. The input unit 12 is used by the user (registrant or searcher) to input the concept index indexAb such as the keyword KW when registering the image data Di in the image database 200 or searching the image database Di for the image data Di. Is what you use.

The display unit 11 and the input unit 12 are also provided.
The operation is controlled by the display control unit 14. The image data reading unit 13 is used to capture image data Di of an image to be registered or used as a basis for search. The image data reading unit 13 is composed of a scanner or the like having an image data reading function.

Image data Di of the image to be registered or the basis of the search is supplied from the image data reading unit 13 or the like to the conceptual index creating means 15, and
The keyword KW or the like for the image data Di input by the user is supplied from the input unit 12 described above. When the image data Di to be registered is supplied, the conceptual index creating means 15 extracts the feature amount Q from the image data Di, and based on the feature amount Q, the color of the image data Di,
A conceptual index indexAa such as orientation, position, etc. is created.

Further, the concept index creating means 15 is
When there is a keyword KW or the like for the image data Di input by the user from the input unit 12, a concept index indexA including the keyword KW or the like.
Also create b. Then, the concept index creating means 15
Is the concept index indexAa created from the feature quantity Q of the image data Di and the concept index indexAb composed of the keyword KW and the like.
As A, it is supplied to the concrete index creating means 16.

The concrete index creating means 16 retrieves the concrete index indexXb from the image database 200 in which various registered image data D are accumulated based on the concept index indexA supplied from the concept index creating means 15. . Then, the specific index creating means 16 extracts and combines the specific indexes indexXb included in the retrieved one or more image data Dx to determine the credibility Cx of each specific index indexXb.

Further, the concrete index creating means 16
Connects the specific index indexXb with the determination of the credibility Cx to the above-mentioned conceptual index indexA of the image data Di to be registered, and the registration index creating means 17 together with the image data Di.
Supply to. The registration index creating means 17 connects the data via the metadata M on the internet 100 or the internet 100 based on each concrete index indexXb for which the credit level Cx supplied from the concrete index creating means 16 is determined. A specific index in the possible information data I of the dictionary database 400
Search for dexYb.

The registration index creating means 17
Extracts and combines the specific indexes indexYb possessed by one or more of these searched data Dy to determine the credibility Cy of each specific index indexYb. Further, the registration index creating means 17 extracts the specific index indexZb for searching the image data Di to be registered in the image database 200 from the extracted and combined specific index indexYb based on the determined reliability Cy. Selected, and the selected result is connected to the concept index indexA to obtain a specific index index for storage for searching the image data Di.
Create Zb.

In other words, the registration index creating means 1
When creating the specific index indexZb for storage, 7 retrieves the information data I of the dictionary database 400 that can be data-connected via the Internet 100 and the metadata M on the Internet 100, and the concept index creating means 15 The concrete index indexYb concerning the meaning and related matters of the created conceptual index indexA is acquired, and this concrete index i
A specific index indexZb for searching the image data Di is created based on ndexYb.

The category classification creating means 18 is a specific index indexZb of the image data Di to be registered, which is created by the registration index creating means 17.
And a credit C attached to the category, a category is created, and based on this category, the image data Di is automatically classified together with similar images in the image database 200, and the result is stored in the image category database 300. Register for renewal.

The image data searching means 19 includes (1) the concept index in created by the concept index creating means 15.
dexA, (2) Specific index indexZb for search (that is, for storage in the image database 200) created by the registration index creating means 17, (3) Category created by the category creating means 18, search using each Provide the method to users (searchers).

By the way, in the image database creation / retrieval system 1 of the present embodiment, the above-mentioned display unit 11, input unit 12, image data reading unit 13, and display control unit 14 are described.
Adopts a client / server system configuration. As a result, a plurality of users can use the system 1 via the network.

In the image database creation / retrieval system 1 of the present embodiment, the image to be registered or the basis of the retrieval is a still image of any format, a still image such as a document input by a scanner, Targets are moving images shot by video cameras and digital cameras, broadcast programs stored on broadcasting stations and home video servers, content data such as movies, and moving images shot by surveillance cameras and WWW cameras. ing.

FIG. 2 is a "swallow" photograph 500 as an example of an image that is to be registered in the image database 200 or is a basis for searching the image database 200. Next, in the image database creating / searching system 1 of the present embodiment, a case where the image data Di is to be registered in the image database 200 will be described by taking the "swallow" photograph 500 as an example.

FIG. 3 shows an image database 200 in the image database creation / retrieval system 1 of this embodiment.
5 is a flowchart of a registration process to the. The image database creation / retrieval system 1 includes an image data reading unit 13
From the above, the image data Di to be registered is acquired.
The image database creation / retrieval system 1 also includes the keyword KW or the like input by the user (registrant) from the input unit 12 as necessary in association with the image data Di to be registered. To obtain (step S31).

The acquired image data Di is first supplied to the concept index creating means 15 to generate the image data D.
The feature value Q of i is extracted (step S32). Examples of the feature amount Q of the image data Di extracted by the concept index creating means 15 include the color, direction, position, etc. of the image data Di to be registered. Then, the concept index creating means 15 uses the concept index indexA (indexAa and indexA) based on the extracted feature amount Q of the image data Di and the keyword KW that is appropriately input.
b) is created (step S33). Regarding the creation of the conceptual index indexA, the conceptual index creating means 15 first uses the feature quantities Q such as color, orientation, position, etc. of the image data Di to be registered, respectively, to create the conceptual index of the image data Di to be registered. inde
Create xAa.

For example, the concept index creating means 15
In the case of the "swallow" photograph 500 shown in FIG.
From the feature quantities Q1, Q2, and Q3 relating to the color, direction, and position extracted as the feature quantity Q of the image data Di, for example, "black" is used as the concept index indexAa1 relating to color, and the concept index indexAa2 relating to direction.
As "leftward", also a conceptual index about position
A conceptual index indexAa consisting of "center (image center)" is created as indexAa3.

Further, the concept index creating means 15
In addition to the conceptual index indexAa created based on the feature quantity Q of the image data Di, the user operates the input unit 12 or the like to input a keyword KW or the like for more detailed search. In this case, the concept index indexAb composed of this keyword KW and the like is also created.

Here, the user (registrant) of the "swallow" photograph 500 shown in FIG. 2 corresponds to the "swallow" photograph 500, for example, "bird", "swallow", "Siebold collection". Keyword K such as "," migratory bird
It is assumed that W1, KW2, KW3, and KW4 are input.
In this case, the concept index creating means 15 is a concept index indexAb1, indexAb2, index composed of these keywords KW1, KW2, KW3, KW4.
Create Ab3 and index Ab4.

FIG. 4 is a photograph 5 of the "swallow" created by the concept index creating means 15 in this way.
It is explanatory drawing of the conceptual index indexA which made 00 an example. As shown in FIG. 4, the concept index indexA is
It is composed of a concept index indexAa created based on the feature amount Q of the image data Di of the image to be registered, and a concept index indexAb composed of the keyword KW or the like directly input by the user.

With respect to the above-mentioned registration of the image data Di, with respect to the concept index indexAb consisting of the keyword KW etc. directly input by the user, the concept index creating means 15 gives each of the concept indexes indexAb an identifier ident. The concept index indexAa created based on the image data Di of the image to be registered is distinguished from the concept index indexAa.

As a result, this concept index index
The keywords KW and the like that compose Ab can be used also as specific indexes indexXb such as the meaning of the concept index indexAa and related matters by the assigned identifier ident. The image data Di to be registered is displayed on the Internet 100 via the network by using the display unit 11, the input unit 12, and the display control unit 14.
It is possible to construct the screen created in 1. and the screen created in SGML, XML, etc. with the same configuration.

FIG. 5 shows the Internet 1 as an example.
00 is a screen 600 created in HTML published on the Internet. FIG. 6 shows the Internet 1 shown in FIG.
10 is an explanatory diagram of definition information t of a screen 600 created in HTML published on 00. FIG. For example, when the image data Di to be registered is a Web (World Wide Web) screen, the UR of the Web screen 600 is input from the input unit 12.
Input L (uniform resource locator). Based on this, the concept index creating means 15 collects the corresponding Web screen 600 from the URL. And HTML
Since the definition information t of the screen 600 created in step 1 is distinguished by the tags “<” and “>” with respect to the concept index indexAa extracted from the image data Di, the concept index creating unit 15 uses each tag. The concept index indexAc can be cut out for each.

FIG. 7 is a diagram showing an example in which the concept index indexAc is cut out for each tag "<", ">" from the HTML definition information t. The concept index indexAc cut out from the HTML definition information t has the identifier ident, as in the case of the concept index indexAb composed of the keyword KW or the like input by the user (registrant). Let
Concept index indexA extracted from image data Di
Discriminate against a. This allows the concept index in
dexAc can also be used as a concrete index indexXb such as the meaning of the concept index indexAa and related matters.

Then, returning to FIG. 3, the concrete index creating means 16 creates the concept index index created by the concept index creating means 15 shown in step S33.
Whether or not there is image data Dx similar to the image data Di to be registered in the image database 200 in which various image data D are accumulated based on A (that is, indexAa, indexAb, indexAc). A search is performed (step S34).

In the present embodiment, the search process of step S34 is performed by the concept index inde of the image data Di to be registered, as described with reference to FIGS.
The concept index indexAb by inputting a keyword having an identifier ident to xA and the HTML definition information t
If there is a concept index indexAc cut out from the specific index creating means 16 with the concept indexes indexAb and indexAc, the image database 2
The specific index indexDb of each image data D stored in 00 is searched.

Then, in this case, the concrete index creating means 16 uses the concept index indexAa having no identifier ident to generate a concept index for each image data D stored in the image database 200.
The indexDa is not normally searched. That is, the concrete index creating means 16 identifies the concept index indexA if there are concept indexes indexAb and indexAc having the identifier ident.
The concept index indexDa of each image data D stored in the image database 200 with the concept index indexAa having no ident is not searched.

As a result, in the image database creation / retrieval system 1 of the present embodiment, it is possible to minimize unnecessary retrieval of the specific index indexXb, and to speed up and improve the efficiency of the registration process itself. ing. Further, regarding the image data Di to be registered, the user does not directly input the keyword KW or the like from the input unit 12, and the concept index indexA is only the concept index indexAa acquired from the feature amount Q of the image data Di. Even when configured, the concrete index creating means 16 is adapted to correspond. In this case, the concrete index creating means 16
Searches for the conceptual index indexDa of each image data D stored in the image database 200 with this conceptual index indexAa, and acquires the corresponding specific index indexDb.

In the above description, the image database 200 searched by the concrete index creating means 16 is used.
Need not be in the image database creation / retrieval system 1 itself, but may be another image database or website existing on the network. In this case, the image database creation / retrieval system 1 uses the record numbers or URs added to the other image databases or the plurality of websites existing on the network.
Manage by L. Further, the registered image data D stored in each image database or each website is also a concept index index of each registered image data D.
The image database creation / retrieval system 1 manages Da, a concrete index indexDb of the meaning and related items of the conceptual index indexDa, and a category described later.

As described above, the concrete index creating means 16 searches the registered image data D in the image database 200 for the image data Di to be registered, and the similar registered image data Dx has the specific image data Di. The dynamic index indexXb is extracted (step S34 in FIG. 3).

FIG. 8 shows the image data Di to be registered.
With respect to the specific index indexD of each of the image data D stored in the image database 200, the specific index creating means 16 based on the conceptual index indexAb input by the user (registrant).
It is explanatory drawing which searches b and acquires the concrete index indexXb of the similar registered image data Dx.

First, the concrete index creating means 16
Is the user for the image data Di to be registered.
Concept index index entered by (registrant) as a keyword
Registered image data D (= D1 to Dp, p is the number of registered image data) stored in the image database 200 based on Ab.
The specific index indexDb of is searched. And
The concrete index creating means 16 uses the registered image data D.
From 1 to Dp, concrete indexes indexX11b, indexX22b, indexX31b having the same contents as the conceptual index indexAb of the image data Di to be registered, ...
A plurality of registered image data Dx1, Dx2, Dx3,
... is extracted.

As a result, the image data Di to be registered has the concept index indexAb possessed by the extracted registered image data Dx1, Dx2, Dx3, ... In addition to the concept index indexA (that is, indexAb) possessed by itself. And another concrete index indexX12
b, indexX21b, indexX32b, ... Are acquired. That is, the image data Di to be registered is obtained from the extracted registered image data Dx1, Dx2, Dx3, ... Based on the specific indexes indexXb1, indexXb2, index.
xXb3, indexXb4, ... Are acquired.

FIG. 9 shows the acquired concrete indexes indexXb1, indexXb2, indexXb3, ... In the concrete index extraction processing shown in step S34 in FIG.
.. It is explanatory drawing regarding the determination process of each credit Cx. The concrete index creating means 16 receives the concept indexes indexAb1, indexAb2 from the concept index creating means 15 for the image data Di to be registered.
The specific index indexDb of the registration data D in the image database 200 is searched based on the “concept index 1” and the “concept index 3” supplied as

It is assumed that the "concept index 1" and "concept index 3" are both keywords KW input by the user (registrant) from the input unit 12. Here, since the concept index indexAb can also be used as the concrete index indexXb as described above, the “concept index 1” and the “concept index 3” themselves are substantially the same. The description will be made assuming that the specific contents are the same as “specific index 1” and specific index 3 ”.

Then, the concrete index creating means 16
From the image data D1 to Dp already registered in the image database 200,
3 (concrete index 1, 3) ", it is assumed that image data Dx1, Dx2, Dx3 having at least one of" concrete index 1, 3 ". In addition, the image data Dx2 has "specific indexes 2, 4, 5, ..."
In addition to "specific indexes 1, 3", "specific indexes 5, 6, 7, ..."
Has "specific indexes 6, 7, 8, 9, ..." In addition to "specific index 1".

Therefore, the image data Di to be registered is not limited to the "specific indexes 1, 3" created by the conceptual index creating means 15, and the "specific index 2," is searched by searching the image database 200.
, 4, 5, 6, 7, 8, 9, ... "Is acquired as a specific index indexXb.

Next, the concrete index creating means 16
Is the acquired “specific index 1, 2, 3, 4,
5, 6, 7, 8, 9, ... ", the image data Di to be registered and the searched image data Dx1,
The degree of overlap W between Dx2 and Dx3 is calculated, and "specific index 1, 2, 3, 3" is calculated according to the value of the degree of overlap W.
The credibility Cx for each of 4, 5, 6, 7, 8, 9, ...

Specifically, for example, the concrete index creating means 16 firstly registers the image data D to be registered.
i is created by the concept index creating means 15,
The image of the first search result with the “concept index 1 (that is, concrete index 1)” and “concept index 3 (that is, concrete index 3)” of the concept index indexAb that became the searcher this time as the matching source "Specific index 1", "specific index 2", "specific index 3", "specific index 4", "specific index 5" of data Dx1 ...
As the matching destination, and between the matching source and the matching destination, concrete indexes (indexAb and in
dexX1b).

When executing this first matching process, the "specific index 1" of the matching source,
An initial value (= 1) of the degree of overlap W is set in association with each "specific index 3". By executing this matching process, the specific index indexAb of the matching source and the specific index index of the matching destination.
If there is a specific index with the same contents as X1b,
Concrete index of the same content of the matching source index
The overlap degree W corresponding to Ab is incremented by 1.

As a result, in the case shown in FIG. 9, the overlapping degree W of the "concrete index 1" and the "concrete index 3" of the matching source is counted up to "2".
Further, if any of the concrete indexes indexAbm (where m is a positive integer) in the concrete index indexAb of the matching source is not in the concrete index indexX1b of the matching destination, it is in the matching source. The degree of overlap W corresponding to the specific index indexAbm is 1
It has not been counted up, it remains as it is.

Further, the specific index of the matching destination in
Any specific index inde in dexX1b
xX1mb is the specific index indexAb of the matching source
If it does not exist, the concrete index indexX1mb of this matching destination is the concrete index index of the matching source.
Added to Ab and the added specific index index
An initial value (= 1) of the degree of overlap W is set in correspondence with X1mb.

In this way, in the case shown in FIG. 9, the specific index indexAb of the matching source has the existing "specific index 1" and "specific index 3" as the result of this matching process. In addition, specific indexes indexX12b, indexX14b, indexX15b having the contents of "specific index 2", "specific index 4", "specific index 5" ... Are added.

For the overlapping "concrete index 1" and "concrete index 3", the overlapping degree W is incremented by 1 to "2", and
For each of the added "specific index 2", "specific index 4", "specific index 5", ..., The initial value (= 1) of the degree of overlap W is set.

In this way, first, of the concept indexes indexA of the image data Di to be registered, the concept indexes indexAb1 and indexAb2 that can also be used as concrete indexes are set as the matching source, and the first search result Specific index i of image data Dx1
The specific index creating means 16 completes the matching process between the specific indexes of the two with ndexX1b as the matching destination.

Next, the concrete index creating means 1
The reference numeral 6 indicates the result of the immediately preceding matching process for each of the remaining image data Dx2 to Dxn until the image data Dxn (where n ≦ p) as a search result for which the matching process is not completed is exhausted. And the specific index in any of the remaining image data Dx2 to Dxn in
The specific index matching process with dexX2b to indexXnb as the matching destinations is repeated.

As a result, the retrieved image data Dx1
~ Dxn All specific indexes indexX1b ~ indexX
For nb, the degree of overlap W is calculated. Then, each time this matching is performed, if there is a "concrete index" having the same content, the degree of overlap W of the "concrete index" having the same content is incremented by one.

In FIG. 9, in the list L, “specific indexes 1 to 9” acquired by the image data Di to be registered and obtained by this matching process are arranged in the descending order of the degree of overlap W as an example. It was done. For example, in the list L, in the case of "specific index 1", since the image data Di to be registered and the searched three image data Dx1, Dx2, Dx3 are provided, the degree of overlap W thereof is W.
Becomes "4". In the case of "concrete index 3", since the image data Di and the retrieved two image data Dx1 and Dx2 have "concrete index 3", the degree of overlap W thereof is "3". , W in the case of "concrete indexes 5, 6, 7"
Becomes “2”, while “specific index 4, 8,
In the case of “9, ...”, the degree of overlap W remains “1 (that is, the initial value)”.

Then, the concrete index creating means 1
6 determines the credibility Cx of the specific index based on the degree of overlap W of each acquired specific index Xb. The concrete index creating means 16 determines the concrete index inde in determining the credibility Cx.
The larger the value of the degree of overlap W of xXb, the more specific index indexXb is the specific index indexXb.
Specific index indexX with a lower degree of overlap W than
It is determined that the credibility Cx is relatively higher than that of b.

For example, in the case of "concrete index 1", the degree of overlap W is "4", which is larger than the degree of overlap W of other "concrete indexes 2-8, ..." Cx is "1", which means that the credibility Cx is the highest. Then, the concrete index creating means 16
Is the plural image data Dx1 to Dx thus obtained.
The specific indexes indexX1b to Xnb of xn are extracted and combined to generate specific indexes indexXb1 to Xb9 in which the credibility Cx is determined for each “specific index of the same content” as shown in list L.

Then, as shown in step S35 in FIG. 3, the specific index creating means 16 determines the specific index indexXb (inde) for which the credibility Cx is determined.
xXb1 to Xb9, ...) Are connected to the image data Di and the concept index indexAa created based on the feature amount Q of the image data Di. In the search of the image database 200 by the specific index creating means 16, the concept index indexA of the image data Di to be registered is the feature amount Q of the image data Da.
Even if there is only the concept index indexAa acquired from
The image database 200 is searched based on this conceptual index indexAa, and the specific index indexXb for which the credibility Cx is determined can be acquired in the same manner.

Here, in the case of this example, in the concrete index indexXb for which the credibility Cx is determined, the credibility of the “concept index 1” and the “concept index 3” of the concept index indexAb which is the current searcher is , A specific index in extracted from the image database 200 by the “concept index 1” and the “concept index 3” as shown in list L of FIG.
dexXb, that is, "specific index 2", "specific index 4" to "specific index 9", ...
・ It is higher than any credit Cx.

This means that for the image data Di to be registered, the keyword KW input by the user (registrant) who is the basis of “concept index 1, 3” from the input unit 12 is standardized in the image database 200. It is shown that the keyword KW has been used.

On the contrary, a plurality of keywords KW are input from the input unit 12 in this way, and a plurality of concept indexes indexAb (for example, i
ndexAb1, indexAb2) Even if generated, if there is a specific index indexXb extracted from the image database 200, which has a higher credibility Cx than any of the concept indexes indexAb1, indexAb2,
With the specific index indexXb, the creditworthiness C
Either the concept index indexAb1 or indexAb2 having a low x can acquire the keyword KW that is standardized and used in the image database 200.

In FIG. 10, the registration index creating means 17 uses the specific index indexXb for which the credibility Cx is determined and the metadata M on the Internet 100.
FIG. 8 is an explanatory diagram for searching the information data I of the dictionary database 400. As shown in step S36 of FIG. 3, the registration index creating means 17 is supplied by being linked to the image data Di from the concrete index creating means 16.
Specific index indexXb for which the credibility Cx is judged
(indexXb1, indexXb2, indexXb3, indexXb4,
...) is searched based on the metadata M on the Internet 100 and the information data I in the dictionary database 400.

In the case of this embodiment, the dictionary database 4
00 does not need to be present in the image database creation / retrieval system 1 itself, and a specific index creation means 1 can be created from the dictionary databases 400 existing on the network.
Based on the concrete index indexXb obtained in No. 6, the index in of the concrete index indexXb (that is, the concept index indexA) and the related matters, etc.
Search for dexYb.

Therefore, the registration index creating means 1
7, like the search process of the concrete index creating means 16, first, from the metadata M on the Internet 100 and the information data I of the dictionary database 400, concrete indexes indexXb1, indexXb2, indexXb.
Information data Dy1, Dy2, D including 3, ...
Search for y3, ...

Here, the registration index creating means 17
Is a specific index indexXb1, indexXb2, inde
xXb3, indexXb4, ... From the information data Dy1, Dy2, Dy3, ... As a result of searching for each, one or more specific indexes indexY1b (indexY11
b, indexY12b, indexY13b, indexY14b, ...
・), IndexY2b (indexY21b, indexY22b, indexY2
3b, indexY24b, ...), indexY3b (indexY31
b, indexY32b, indexY33b, indexY34b, ...
・), IndexY4b (indexY41b, indexY42b, indexY4
3b, indexY44b, ...), ...

For example, if the user has a concept index index
The keywords of “swallow” and “migratory bird” (see FIG. 4) directly input as Ab are specific index creating means 16
When the credibility Cx is determined by and is supplied to the registration index creating means 17 as the specific indexes indexXb1 and indexXb2, the registration index creating means 17 determines the specific indexes indexXb1 and indexXb2 such as “swallow” and “migratory bird”. , Metadata M and dictionary database 4 on the Internet 100
The information data I of 00 is searched.

FIG. 11 shows the search results from the metadata M on the Internet 100 and the information data I of the dictionary database 400 for the specific indexes indexXb1 and indexXb2 of "swallow" and "migratory bird". As a result, for example, the concrete index in
Related information (concrete index indexY13b), etc. regarding the academic classification (concrete index indexY11b), the meaning of the term (concrete index indexY12b), etc., regarding the "swallow" of dexXb1 (concrete indexY13b) Index indexY1b), concrete index indexX
Regarding the "migratory bird" of b2, an example is shown in which related information (specific index indexY2b) such as the meaning of the term (specific index indexY11b), ...

FIG. 12 shows the image data D to be registered.
For i, the registration index creating means 17 searches the metadata M on the Internet 100 or the information data I in the dictionary database 400 based on the specific index indexXb acquired by the specific index creating means 16, and the search result From concrete index Inde
FIG. 7 is an explanatory diagram showing that xY1b, indexY2b, ... Are acquired, and the credibility Cy is calculated based on the overlapping degree W.

First, the registration index creating means 17
As shown in step S36 of FIG. 3, the specific index indexXb (indexXb1, indexXb2, indexXb3, in created by the specific index creating means 16).
For each dexXb4, ...), the metadata M on the Internet 100 and the information data I in the dictionary database 400 are searched.

Then, the registration index creating means 17
Is a specific index indexX from the metadata M and the information data I in the dictionary database 400.
b (indexXb1, indexXb2, indexXb3, indexXb
All the data Dy (Dy1, Dy2, Dy3, Dy4, ...) Including 4 ...
Of the specific index indexYb (indexY1b, index
Y2b, indexY3b, indexY4b, ...) are acquired.

For example, a concrete index indexY1b
Is extracted by searching the metadata M or the information data I in the dictionary database 400 based on “specific index 1” as the specific index indexXb1,
“Index 1” as the concrete index indexY11b, “index 3” as the concrete index indexY12b, “index 4” as the concrete index indexY13b, concrete index indexY1
“Index 5” as 4b, concrete index i
It consists of "index 8" as ndexY15b. Then, any of the specific indexes indexY11b to indexY14b is "specific index 1".

In this way, the registration index creating means 17 uses the other "specific index 3" (specific index indexXb2), "specific index 5".
(Specific index indexXb3), “specific index 7” (specific index indexXb4), ...
Similarly, from the information data I of the dictionary database 400, a specific index indexY2b, a specific index indexY3b, a specific index indexY4b, ...
To obtain.

Then, the registration index creating means 17
Is a specific index indexY1b, indexY2b, indexY3b, index acquired from the metadata M on the Internet 100 or the information data I of the dictionary database 400.
The matching process of Y4b, ... Is performed.

In FIG. 12, this matching process is performed by, for example, first, "index 1, 3, 4, 5, 8" of the specific index index Y1b which is a search result by the "specific index 1" of the specific index index Xb1.
As the matching source, and the specific index indexYb of the specific index indexYb2, which is the search result by the specific index 3 of the specific index indexXb2.
8 "is used as the matching destination, and is performed by the registration index creating means 17 in the same manner as in the case of the specific index creating means 16 described above.

That is, the registration index creating means 17
By both concrete indexes indexY1b, indexY2
b Specific index Index1 included in each
Matching of 1b to Y15b and Index Y21b to Y24b is performed, and the degree of overlap W is obtained for each concrete index having the same content.
Is calculated.

In this matching process, the remaining specific indexes indexY3b, indexY4b, ...
The matching process that uses the result of the matching process of b as the matching source and the remaining specific index IndexYkb as the matching destination is a specific index indexYmb (where m is a positive It is repeated until there are no more integers.

In FIG. 12, the list TL shows the results of this matching processing as the concrete indexes indexYb1, indexYb2, in
It is shown for convenience as dexYb3, indexYb4, .... In this example, the degree of overlap W of “index 3” as the concrete index indexYb1 is “6”, and the degree of overlap W of “index 5, 8” as the concrete indexes indexYb2, indexYb3 is “5”, the concrete index The degree of overlap W of "index 4, 7, 9" as indexYb4, indexYb5, indexYb6 is "4", and specific indexes indexYb7, indexY
"Index 1, 2, 6" as b8, indexYb9
It is assumed that the respective overlapping degrees W are “3”.

In the present embodiment, the registration index creating means 17 has the credibility C of the concrete index indexYb.
In determining y, a specific index indexYb
The larger the value of the degree of duplication W of the
The index Yb is determined to have a relatively higher credibility Cy than the specific index indexYb having a value of the overlapping degree W lower than that of the specific index indexYb.

Then, the registration index creating means 17
Is the concrete index inde calculated in this way
An image database corresponding to the conceptual index indexAa created based on the image data Di to be registered and the characteristic amount Q such as color, direction, position, etc. of the image data Di to be registered based on the value of the credibility Cy of each xYb. A specific index indexZb of the image data Di stored in 200 is created.

In the present embodiment, the registration index creating means 17 predetermines only the specific index indexYb in which the value of the credibility Cy is less than or equal to a predetermined value, in other words, the value of the degree of overlap W. Only a specific index indexYb that is equal to or greater than a predetermined value is a specific index for registering a database of image data Di to be registered.
It is set as index Zb.

In the example shown in FIG. 12, the predetermined value is such that the credibility Cy is "4" or less (the duplication degree W is "3").
Above). Therefore, in the example illustrated in FIG. 12, among the specific indexes indexYb, “index 3, specific indexes indexYb1 to Yb9” are stored.
5,8,4,7,9,1,2,6 "will be adopted as the specific index indexZb (indexZb1 to Zb9) for database registration of the image data Di to be registered.

Further, the registration index creating means 17
In addition to the specific index indexZb (indexZb1 to Zb9) for storing the set database, the shooting date, the shooting location, and the image data Di to be registered,
When the additional information indexZc such as the image size and the photographer is input, these additional information are also connected to the specific index indexZb as the specific index indexZc.

As the information to be added, if the image data Di to be registered is a still image, the image data Di
Additional information indexZc such as the shooting date, the shooting location, the image size, the photographer, etc. can be considered. Also, if it is a broadcast program,
Additional information indexZc such as title, broadcast date, broadcast station, performer, broadcast time, etc. can be considered. In the case of content data such as a movie, additional information indexZc such as title, release date, performer, producer, etc. can be considered. Then, the concrete index indexZc is the concrete index ind
When connecting to exZb, "1" is added as the credit Cz.

FIG. 13 is an explanatory diagram of the specific index indexZb created in this way, for database registration, to which the determination of the credibility C is added. In this way, the specific index ind with the additional information indexZc added
exZb is associated with the image data Di to be registered and the conceptual index indexAa created based on the feature quantity Q such as color, direction, position, etc. of the image data Di, and thus the above-mentioned FIG. As shown in step S37, it is stored in the image database 200.

Further, the category classification creating means 18 supplies the specific index IndexZb for the image data Di to be registered from the registration index creating means 17, as shown in steps S38 and S39 of FIG. On the basis of the credibility C (that is, the degree of duplication W) and the specific index IndexZb, categories are created in the descending order of the credibility C, that is, the degree of duplication W, and the categories are registered in the image database 200. Similar images of the image data D are automatically classified and registered in the image category database 300. Subsequently, in the image database creating / searching system 1 of the present embodiment, the desired image data Do is stored in the image database 20.
A case of trying to search from 0 will be described.

FIG. 14 shows an image search screen 700 displayed on the display unit 11 by the image data search means 19 when searching the image database. In the image database creation / retrieval system 1 of the present embodiment, it is possible to perform a search by using three types of search sections 710: keyword input 711, image data input 712, and URL input 713. In the image search screen 700 shown in FIG. 14, the keyword input 711 is specified, and the input keyword KW is displayed in the input display field 720 formed in the screen 700. There is.

FIG. 15 is a flowchart of the search process performed by the image database creation / search system 1 when the keyword input is selected as the search category 710.
When the search operation is started, the user (searcher) inputs the search keyword KW (step S1501). The search target here is the above-mentioned conceptual index indexA, and the specific index indexZb of the related items such as the meaning of the conceptual index indexA and academic classification.

For example, when "swallow" is input as the search keyword KW in the search key input (step S150
1), the image data search means 19 includes an image database 20.
In the image data D stored in 0, "swallow"
The image data Do having the specific index indexZb is searched (step S1502).

Then, the image data searching means 19 confirms whether or not there is a search hit image data Do.
(Step S1503) If there is a search hit image data Do, it is checked whether or not there are multiple hits (Step S1504). If there is no search hit image data Do, in step S1509 described later. Described Move to concept index search. Here, if there are a plurality of search hit image data Do (Do1, Do2, ...), the image data search means 19 sorts (sorts) the search hit image data Do1, Do2 ,. ) Is performed (step S1505).

In this sorting, the image data searching means 19 uses the search hit image data Do1, Do2,
... Specific index of each "swallow"
Paying attention to the value of the credibility C of the index Zb, the image data Do having a higher credibility C of the "swallow" as the specific index indexZb is more likely to be the image data Do desired by the user (searcher). It is assumed. Therefore, the image data search means 19 sorts (aligns) the search hit image data Do1, Do2, ... By the "swallow" of the specific index indexZb in the search hit image data Do1, Do2 ,. It is performed based on the value of the trustworthiness C that the user has.

The image data search means 19 uses this sort.
The "swallow" similar images of the image data Do1, Do2, ... Are displayed in a list in the descending order of the credibility C of the (aligned) result (step S1506). In displaying this list,
The image data are displayed on the display unit 11 in descending order of reliability. Then, the user selects the image data Do1, Do2, ... From the input unit 12 while looking at the display result on the display unit 11 (step S1507), and the image data search means 19
Displays the selected image data Ds and its index on the display unit 11 (step S1508).

By the way, if the number of hits confirmed in the hit number confirmation processing in step S1504 is one, the image data and the image data are directly displayed without sorting the similar images in the order of credit (step S1506) and displaying the list (step S1507). The index is displayed on the display unit 11 (step S1508).
On the other hand, if the number of hits is 0 in the confirmation processing of the presence or absence of the number of hits in step S1503, the concept index inde
Search for xZa (that is, indexAa) (step S1)
509).

Then, with respect to the search result, it is confirmed whether or not there is a search hit image data Do (step S1510), and if there is a search hit image data Do, it is confirmed whether or not there are a plurality of search hits. Do (step S151
1). As a result, when there are a plurality of image data items (similar image data) Do (Do1, Do2, ...) That have been searched for, similar images that have been searched for are searched, as in the case of the above-described specific index indexZb search. Data Do
1, Do2, ... Are sorted in the order of trustworthiness (step S1505), and a list of similar images is displayed (step S150).
6).

Also in this case, the list display is the image data Do.
1, Do2, ... in descending order of credibility (in this case, the credibility C is, for example, the individual concept indexes indexAa1, indexAa2, ... In the concept index indexZa.
The number of similarities to, specifically, three, such as color, orientation, and position, are matched, any two are matched, or only one is matched.

Based on this, the user looks at the display result on the display unit 11 and inputs the image data Do1,
Do2, ... Is selected and the image data searching means 19 is selected.
Displays the selected image data Ds and its index on the display unit 11 (step S1507). On the other hand, if there is one image data Do as a search result, the image data Do and its index are displayed on the display unit 11 as they are, as in the case of the specific index search described above (step S1508).

On the other hand, in step S1511, even if the search is performed using the concept index indexZa, the search result is 0.
If yes, the search operation ends. Therefore, in the search by this keyword input, as shown in FIG. 11, the image data D registered in the image database 200 is the information collected by the registration index creating means 17 and created as the physical index indexZb. Become.

Accordingly, the image data Ds desired by the user (searcher) is the keyword KW different from the keyword KW (see step S1501) input by the user (searcher) at the time of the search, and the image database 200 by the registrant. The user (searcher) can search the image database 200 for the desired image data Ds even when registered in the. That is, even when the user (searcher) searches for "swallow" by entering "swallow" or "swallow" as keywords, for example, from the information shown in FIG. 11, enter "swallow" as keywords. The same result as when searching with

Further, the color of the image data Di to be registered (for example, "blue black") is the concept index creating means 1
Even if it is not extracted as the conceptual index indexAa in 5 and does not become the specific index indexAb in the specific index creating means 16, it is collected in the meaning of "swallow" by the registered index creating means 17, for example, in FIG. Depending on the information shown, "blue black", "white",
You can search by "reddish brown".

FIG. 16 is a flow chart of a search process performed by the image database creation / search system 1 when a search is performed using an image file as a search key. By starting the search operation, for example, from the input unit 12,
When the file name of the image data Dk is input as the search keyword KW (step S1601), the image data searching means 19 uses the concept index creating means 15 to extract the feature of the image data Dk of the file name, A conceptual index indexAa of the data Dk is created (step S1602). Then, the image data searching means 19 uses the concrete index creating means 16 to create the conceptual index created by the conceptual index creating means 15.
The image database 200 and the dictionary database 400 are searched based on the index Aa (step S1603).

After that, the image data searching means 19 confirms whether or not there is the image data Dp which is a search hit by the concrete index creating means 16 (step S160).
4) If there is a search hit image data Dp, a specific index index of this search hit image data Dp
From Db, the specific index index of the image data Dk
Xb is created (step S1605). Then, the image data search means 19 uses the created specific index.
The image database 200 is searched with the index Xb (step S1606), and it is confirmed whether or not there are a plurality of image data Do obtained as a result of the search (step S1607).

As a result, a plurality of image data Do (Do
1, Do2, ...), the same as in the case of the processing by steps S1505 and S1506 in FIG.
The image data search means 19 sorts the image data Do (that is, the similar image data Do of the image data Dk) in the order of the credibility C (step S1608), and displays a list of similar images by the similar image data Do. To 11
(Step S1609). Here, the list display in this case is displayed in the descending order of the credibility C of the image data Do.

After that, the user sets the image data Do1, Do
2 is selected (step S1610), the image data search means 19 displays the selected image data Ds and its index on the display unit 11 (step S1611),
The search operation ends. On the other hand, if there is one search result, the image data Do and its index are displayed on the display unit 11 as they are, as in the case of the specific index search described above (step S1611). On the other hand, in the search result confirmation processing in step S1604, if there are 0 results, the search operation ends.

Therefore, in the search by key input of this image file, the conceptual index indexAa relating to the image data Dk of the key input image file is created, and further, the specific index indexXb of the image data Dk is created based on this. , This concrete index i
The image data Ds desired by the user (searcher) is searched based on ndexXb.

As a result, the concept index indexA
Even if the desired image data Ds cannot be narrowed down due to a large number of search results when only a is searched, the specific index indexXb created based on the conceptual index indexAa is used to search for the search result. Since the search is further performed, the efficiency for obtaining the desired image data Ds is improved.

FIG. 17 shows U of the Web screen as a search key.
9 is a flowchart of a search process performed by the image database creation / search system 1 when a search is performed by inputting RL. When the search operation is started, for example, the input unit 12
When the URL of the Web screen is input as the search keyword KW from (step S1701), the image data search means 1
9 searches the website from the input URL and collects the definition information t of the web screen (step S170).
2).

Then, the image data searching means 19 confirms whether or not the Web screen of the specified URL exists.
(Step S1703), if the Web screen of the specified URL exists, the conceptual index inde shown in step S1704
While the xAc creation process (cutout process) is performed, if not present, the search operation ends. In the process of creating the conceptual index indexAc in step S1704, as described above, the definition information t of the Web screen created in HTML.
Is the concept index inde extracted from the image data Di
The tags "<" and ">" are distinguished from xAa.

Therefore, the image data searching means 19 cuts out the concept index indexAc by cutting out the concept index indexAc for each tag by the concept index creating means 15. Then, the concept index indexAc cut out from this definition information t is
The concept index creating means 15 has an identifier iden.
The concept index indexAa extracted from the image data Di is differentiated from the concept index index
A concrete index indexX such as meaning of Aa and related matters
It can also be used as b.

Therefore, when searching the image database 200, the image data search means 19 does not search the concept index indexDa of the image data registered in the image database 200 with this concept index indexAc, and Only the index indexDb is searched (step S1705). As a result of this search, the image data search means 19 finds similar image data Do.
Exists (step S1706), and if similar image data Do exists, a plurality of cases (Do1, Do2,
() Is checked (step S1707).

As a result, a plurality of image data Do (Do
1, Do2, ...), the same as in the case of the processing by steps S1505 and S1506 in FIG.
The image data search means 19 sorts the image data Do in the order of the credibility C (step S1708), and displays a list of similar images by the image data Do on the display unit 11 (step S1709). Here, the list display in this case is displayed in the descending order of the credibility C of the image data Do.

After that, the user sets the image data Do1, Do
2, is selected (step S1710), and the image data search means 19 displays the selected image data Ds and its index on the display unit 11 (step S1711),
The search operation ends. On the other hand, if the retrieved result is one, the image data Do and its index are displayed as they are on the display unit 11 (step S1711). In contrast,
If the result of the search result confirmation processing in step S1706 is 0, the search operation ends.

Therefore, even when the search is performed based on the displayed Web screen by inputting the URL, the specific index index cut out from the definition information t of the Web screen is displayed.
Concept index that can also be used as Xb in
With dexAc, desired image data Ds can be retrieved from the image database 200 quickly and efficiently.

FIG. 18 is a diagram showing an example of the image search category screen. Here, the category is created by the category classification creating means 18 based on the specific index indexZb having the high credibility C of the image data Di.
In creating this category, the specific index indexZb of the image data Di is checked for duplication with other categories (specific index indexXb) already registered in the image category database 300, so that the same category is not created. Has been done.

For example, the specific index indexZb of the image data Di is already in the image category database 30.
If the registered category name is 0, the category creating means 18 increments the number of images PC stored corresponding to the registered category name, and the image data Di in the image database 200. The data for identifying is additionally stored in association with the category name.

On the other hand, if the specific index indexZb of the image data Di is not a category name already registered in the image category database 300,
Image data Di was acquired (in this case, credit level is 3 or higher)
For the specific index indexZb, the overlap value with the specific index indexDb of the image data D already registered in the image database 200, that is, the number of images PC having the same specific index indexZb is checked. As a result, if the number of images PC becomes a preset predetermined value, the specific index indexZb is
While being registered as a category name in the category database 300, if the number of images PC is less than a predetermined value, the specific index indexZb is not registered as a category name. Then, when registering this specific index indexZb in the category database 300 as a category name, the same specific index indexZ
Data for specifying the image data Di having b (indexDb) in the image database 200, and the number of previous images P
C is stored correspondingly.

In this case, in the category screen display of the image search, for example, first, from the categories registered in the image category database 300, a list of category names (upper categories) whose number of images PC is larger than a certain size is displayed. The screen thus displayed is displayed on the screen of the display unit 11. Next, when a desired category is selected from these, a screen displaying a list of category names (upper category) having the selected category, having a smaller number of images PC than the previous time, and having a certain size or more is displayed. The surface is displayed on the section 11. Therefore, by repeating such category selection work and performing the same processing as the processing by steps S1505 and S1506 in FIG. 15 at a predetermined stage, desired image data Ds can be quickly and efficiently obtained from the image database 200. You can search.

FIG. 19 shows a case where the above-mentioned keyword input, image data input, or URL input Web screen is used.
6 shows an example of a list display of searched image data (similar image data) Do when searching for desired image data Ds from the image database 200. This is because when the keyword is input, the processing result of step S1506 in FIG. 15 described above is used, when the image data is input, the processing result of step S1609 described above in FIG. 16 is used, or when the Web screen in which the URL is input is used, This corresponds to the processing result of step S1709 in FIG. 16 described above.

Therefore, when the desired image data Ds is searched from the image database 200, if there are a plurality of similar image data Do (Do1, Do2, ...) As the search result, the index indexA used at the time of the search. Based on the above, thumbnail images of each image data Do (Do1, Do2, ...) Are displayed in a list on the screen of the display unit 11. Moreover, the list display is the image data D.
For each o (Do1, Do2, ...), the image data Do (Do1, Do2 ,. Do (Do1, Do2, ...)
Thumbnail images are arranged on the screen and displayed vertically or horizontally, or vertically or horizontally in one row or in one horizontally or horizontally.

Then, the image displayed on the image list screen is
If it is a broadcast program or content data, the frames from which the indexes of those data have been extracted may be displayed. As a result, in the image selection processing in step S1507 in FIG. 15, step S1610 in FIG. 16, or S1710 in FIG.
The desired image data Ds can be easily and efficiently selected by using a pointing device such as a keyboard and a mouse shown in 2.

In the subsequent display of the selected image in step S1508 in FIG. 15, step S1611 in FIG. 16, or S1711 in FIG. 17 described above, the selected image data Ds requires special pure. If it is something, it will start Pure and display it. If the selected image data Ds is a broadcast program or content data, that image is displayed.

[0123]

According to the present invention, the concept index created from the feature amount of the image data to be registered in the image database and the concept index composed of the input keyword correspond to the image data to be registered. When storing the image data in the image database, normalization (standardization) is performed by the specific index assigned to the image data already registered in the image database. It is possible to prevent the search accuracy from deteriorating due to subtle shifts in the index information and to search the image database accurately and efficiently.

Further, the learning effect of the index information added to the image data accumulated in the image database can be expected by normalizing (standardizing) the specific index. Further, according to the present invention, even when an image database search is performed based on image data or a Web screen, a concept index (in this case, a specific index) can be cut out and a search can be performed based on this. Therefore, it is possible to search the image database accurately and efficiently. Furthermore, since the specific index assigned to the image data registered in the image database is assigned a credibility for each specific index, it is possible to search the image database more accurately and efficiently. become.

[Brief description of drawings]

FIG. 1 is a block diagram showing a configuration of an image database creation / search system according to an embodiment of the present invention.

FIG. 2 attempts to register in the image database 200,
Alternatively, it is a “swallow” photograph 500 as an example of an image serving as a basis for searching the image database 200.

FIG. 3 is a flowchart of registration processing in the image database 200 in the image database creation / search system 1 according to the present embodiment.

FIG. 4 is an explanatory diagram of a concept index indexA created by a concept index creating unit 15, taking a photograph 500 of “swallow” as an example.

FIG. 5 is a screen 600 created in HTML published on the Internet 100 as an example.

FIG. 6 HT published on the Internet 100
It is explanatory drawing of the definition information t of the screen 600 produced by ML.

FIG. 7 shows each tag “<”, from the HTML definition information t.
It is a figure which shows the example which cut out the concept index indexAc for every ">".

FIG. 8 is a diagram showing a concrete index creating unit 16 based on a conceptual index indexAb.
FIG. 6 is an explanatory diagram for searching for a specific index indexXb of similar registered image data Dx.

FIG. 9 is a specific index acquired in the specific index extraction processing, and acquired specific indexes indexXb1, indexXb2, index.
It is explanatory drawing regarding the determination process of each credit Cx.

FIG. 10: Specific index for which the credibility Cx is determined
Registration index creation means 17 using indexXb
6 is an explanatory diagram for searching the metadata M on the Internet 100 and the information data I in the dictionary database 400. FIG.

FIG. 11 shows the search results from the metadata M on the Internet 100 and the information data I of the dictionary database 400 for the specific indexes indexXb1 and indexXb2 of “swallow” and “migratory bird”.

FIG. 12 shows image data Di to be registered,
The registration index creating unit 17 uses the specific index in acquired by the specific index creating unit 16.
Based on dexXb, the metadata M on the Internet 100 and the information data I in the dictionary database 400 are searched,
From this search result concrete index Index Y1b, inde
xY2b, ... and credit C based on the degree of duplication W
It is explanatory drawing for calculating y.

FIG. 13 is an explanatory diagram of a specific index indexZb for database registration, to which a determination of the credibility C is added.

FIG. 14 is an image search screen 7 displayed on the display unit 11 by the image search means 19 when searching the image database.
00.

FIG. 15: Image database creation / search system 1 when keyword input is selected as a search category 710
6 is a flowchart of a search process performed by

FIG. 16 is a flowchart of a search process performed by the image database creation / search system 1 when a search is performed using an image file as a search key.

FIG. 17 is a flowchart of search processing performed by the image database creation / search system 1 when a URL of a Web screen is input as a search key and a search is performed.

FIG. 18 is an example of a category screen for image search.

FIG. 19 is the searched image data when the desired image data Ds is searched from the image database 200.
It is an example of a list display of (similar image data) Do.

[Explanation of symbols]

1 Image database creation / search system 11 Display 12 Input section 13 Image data reading unit 14 Display control unit 15 Concept index creation means 16 Concrete index creation means 17 Registration index creation means 18 Category classification creation means 19 Image data retrieval means 100 Internet 200 image database 300 image category database 400 dictionary database 500 photos of "swallows" 600 HTML screen 700 image search screen

Claims (5)

[Claims] 1. A concept index creating means for creating a concept index from the characteristics of image data to be registered in the image database; and a concept index created by the concept index creating means, which has already been registered in the image database. Specific index creating means for searching the image data, extracting a specific index included in the searched image data, and determining the credibility of the extracted specific index, and the credibility determined by the specific index creating means Based on the specific index marked with, a database other than the image database is searched, a specific index included in the searched data is extracted, and the credibility of each extracted specific index is determined. Based on the judgment result, the image to be registered Image database creation system, characterized in that in correspondence with the data and a register indexing means for setting a specific index to be stored in the image database. 2. The image database creating system according to claim 1, wherein the databases other than the image database are dictionary databases connected via a network or metadata on the Internet. 3. The credibility determination by the specific index creating unit and the registration index creating unit is performed based on the degree of overlap of specific indexes of the searched data among the searched data. The image database creation system according to claim 1 or 2. 4. The registration index creating means combines the credibility of the set specific index in outputting the specific index set corresponding to the image data to be registered for registration in the image database. The output is carried out as follows.
The image database creation system according to 2 or 3. 5. A concept index creating means for creating a concept index from the characteristics of image data to be registered in the image database; and a concept index created by the concept index creating means, which has already been registered in the image database. Specific index creating means for searching the image data, extracting a specific index included in the searched image data, and determining the credibility of the extracted specific index, and the credibility determined by the specific index creating means Based on the specific index marked with, a database other than the image database is searched, a specific index included in the searched data is extracted, and the credibility of each extracted specific index is determined. Based on the judgment result, the image to be registered From the image database created by the image database creation system comprising a registration indexing means for setting a specific index for storing in correspondence with the data to the image database, a keyword input, image data input, U
An image database search system, wherein a desired image is searched by any one of three search categories of RL input.