JPH09138804A - Picture retrieving device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To automatically obtain a picture data base from picture data scattered on a network. SOLUTION: A picture information input part 1 collect-inputs picture data on the network and its auxiliary information and a feature of a picture extraction part 8 extracts the feature of a picture based on the method stored in a picture feature extracting method describing memory 9. A picture category detection part 11 detects a picture category based on a method stored in a picture category detection method describing memory 12. Thereby, a retrieving value to the feature of the picture, a retrieving value to the picture category and a retrieving value of picture auxiliary information are set to retrieve and to display the result.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image search apparatus, and more particularly, to an image search apparatus for an image database in searching resources on a network.

[0002]

2. Description of the Related Art As a conventional search device on a network, for example, there is a category classification service on the Internet. This classifies each homepage into categories according to the content of the title and human resources. This allows the user to obtain the network address of the desired information.
However, what uses the contents and characteristics of the image as the search means is not known at present. As a device for retrieving an image, a device is known in which a keyword manually given to each image when a database is created is used and a search is performed using this. However, because of the enormous amount of effort required to create a database, for example, when performing a search using image feature amounts, several types of image feature amounts are extracted when the database is created, and the feature amounts are similar to the specified image. Japanese Patent Application Laid-Open No. 2-48772 which searches for an image that is present, or an individual user associates a sample image with sensory data such as "bright" to associate it with an image feature amount,
Japanese Patent Laid-Open No. 5-120353 and the like, which perform a search based on individual senses, are known.

[0003]

However, the above-mentioned prior art has the following problems when used as a search device on a network.

First, for example, in the above-mentioned Japanese Patent Laid-Open No. 2-48772, several kinds of image feature quantities are extracted when a database is created,
An image having a feature amount similar to that of the designated image is searched.
For this reason, there is a problem that the search user side cannot perform similarity search with a new scale after creating the database.

Further, in the above-mentioned Japanese Patent Laid-Open No. 5-120353,
Individual users add sensory data such as “bright” to the sample image to associate the sample image with the image feature amount, and search by individual sensory is possible. Therefore, it is necessary to properly select the sample image, and the mapping from the individual user's sense data in the sample image to the image feature amount is used, which cannot cope with the change in the user's sense. I have a problem.

The present invention has been made in view of the above circumstances, and automatically constructs an image database from images scattered on the network, enables access to the network through the images, and performs image search. Responding to search requests based on diverse values of a large number of users,
Another object of the present invention is to provide an image data search device on a network in which individual users can expand the search system in consideration of search results and other users can use it.

[0007]

According to a first aspect of the present invention, in an apparatus for retrieving image data on a network, an image information input means for collecting the image data on the network and auxiliary information accompanying the image data. Image data storage means for storing the image data, image auxiliary information storage means for storing the image auxiliary information, and image feature extraction means for extracting image features from the image data stored in the image data storage means. An image feature storing means for storing the image features obtained by the image feature extracting means, an image category obtaining means for obtaining an image category from the image features stored in the image feature storing means, and the image category obtaining means. Image category storage means for storing the obtained image category, and image category stored by the image category storage means And the image category reading means for reading,
Search image feature setting means for setting a search value for the image feature, search image category setting means for setting a search value for the image category, search image auxiliary information setting means for setting a search value for the image auxiliary information, Search means for performing a search based on the search value for the set image feature, the search value for the image category, or the search value for the image auxiliary information, and display means for displaying the search result by the search means. This is an image search device characterized by the above.

According to this structure, the image information inputting means collects the image data on the network and the auxiliary information associated therewith, the image feature extracting means obtains the image features from the image data, and the image category detecting means allows the image features to be obtained. The image category is obtained from the image features. Thereby, an image database can be automatically constructed from image data scattered on the network. Further, the search value setting means sets a search range for the image feature, the image category and the image auxiliary information, the search means performs a search for the set search value, and the display means displays the search result to the user. The user can obtain desired image data.

According to a second aspect, the image retrieval apparatus of the present invention is provided with image category retrieval method storage means for storing a plurality of retrieval methods for each image category, and a plurality of image category retrieval method storage means are stored based on the stored retrieval methods. You can search image data.

According to this structure, the image category search method can be stored and the stored image category search method can be read out, and a plurality of image category search methods can be added or changed. Can be used.

According to the third aspect, the image retrieval apparatus of the present invention is provided with the image category retrieval method registration means for registering the image category retrieval method, and the image category retrieval method can be arbitrarily registered.

According to this structure, the user can register the image category search method.

According to a fourth aspect of the present invention, the used search method registration means for registering the used search method and the used search method registered in the used search method registration means are registered in the image search device of the present invention. A used search method reading means for reading the search method is provided.

According to this structure, the used search method is registered, the used search method storage means stores the registered search method, and the search method read means can read the stored search method. It is possible to refer to the history of search methods.

According to a fifth aspect of the present invention, in the image search device of the present invention, an image category name registration means for registering an image category name in the search method, an image category obtained by the image category name registration means and its An image category search method registration means for registering a search method is provided.

According to this structure, the image category name is registered in the search method, and the image category name and the search method are registered. Therefore, a search method with a desired result can be newly registered as an image category.

According to a sixth aspect of the present invention, the image retrieving apparatus of the present invention includes an image feature extraction method storage means for storing an image feature extraction method for each image feature, and each image feature extraction from the image feature extraction method storage means. An image feature extraction method reading means for reading the method is provided.

According to this structure, the image feature extraction method for each image feature can be stored and each image feature extraction method can be read out, and a plurality of image feature extraction methods can be added or changed. , Can be used.

According to the seventh aspect, the image retrieval apparatus of the present invention is provided with the image feature extraction method registration means for registering the image feature extraction method, and the image category extraction method can be arbitrarily registered.

According to this structure, the user can register the image feature extraction method. Claim 8
According to this, the image search device of the present invention is provided with the image information link means for associating the searched image data with the image information, and the image information corresponding to the searched image data is displayed on the display means.

According to this structure, it is possible to display the image information by associating the retrieved image data with the image information, and it is possible to retrieve the image and obtain the accompanying image information at the same time.

Further, according to claim 9, the image retrieving apparatus of the present invention is similar to the similarity measuring means between the retrieved image data and each image data, and the similarity is higher than the similarity obtained by the similarity measuring means. A similar search means for searching is provided.

According to this structure, the degree of similarity between the retrieved image data and each image data can be obtained, and similar images can be obtained.

[0024]

BEST MODE FOR CARRYING OUT THE INVENTION The present invention will be described in detail below based on the embodiments shown in the drawings. The present invention is not limited to this.

FIG. 1 is a block diagram for explaining an embodiment of an image data search device on a network according to the present invention. An image information input unit 1 reads a network address memory storing a network address 2 and a data transfer unit 3 collects and inputs image data and auxiliary information thereof on the network. 4
Is an image database construction unit, which constructs an image database for retrieval. An image auxiliary information memory 5 stores the input image auxiliary information. An image data memory 6 stores the input image data. An image feature extraction unit 8 extracts image features based on the method stored in the image feature extraction method description memory 9. An image feature amount memory 7 stores the image features extracted by the image feature extraction unit 8. An image category detection unit 11 detects an image category based on the method stored in the image category detection method description memory 12. An image category memory 10 stores the image category detected by the image category detecting unit. Reference numeral 13 denotes an index memory, which stores the addresses of 5, 6, 7 and 10 of each image information in the memory.

An image retrieval unit 14 retrieves image data. Reference numeral 15 denotes a matching image search unit that extracts image data matching a search range, which will be described later, from the database. Reference numeral 16 is a matching image memory, which is a matching image searching unit 15
The image data extracted by is stored. Reference numeral 17 is a used search method description memory, which stores the used search method and search value. An image feature search setting unit 18 sets a search range of image features. An image category search setting unit 19 sets a search range of image categories. An image auxiliary information search setting unit 20 sets a search range of the image auxiliary information. A similar image search setting unit 21 sets a similar search range for a given image. An image information link unit 22 extracts image information of each image in the compatible image memory 16 from the database.

Reference numeral 23 is a user interface section,
Input / output of searches and various registrations. 24-28 are user input parts. The user input unit 24 inputs registration of the image category detection method. The user input unit 25 inputs an image feature search. The user input unit 26 inputs an image category search. The user input unit 27 inputs an image auxiliary information search. The user input unit 28 performs registration input of the image feature extraction method. An image display unit 29 displays each image in the compatible image memory 16. An image selection unit 30 selects an image from the image display unit 29. 32
Is a process selection unit, and selects a process for the image selected by the image selection unit 30. The image information display unit 29 displays the image information link unit 2 for the image selected by the image selection unit 30.
The image information obtained in 2 is displayed.

A classification method registration unit 33 registers a classification method of image data. An image feature extraction method registration unit 34 registers the image feature extraction method in the image feature extraction method description memory 9. An image category detection method registration unit 35 registers the image category detection method in the image category detection method description memory 12.

Here, as an example of the image data collection method on the network, the World on the Internet is used.
Wide Web will be explained. World Wid
In eWeb, HTML (Hyper Text)
A hypertext (hereinafter referred to as an HTML file) described in a language called Markup Language is used for document expression, and as shown in FIG.
N (Universal Resource Locat
or), which forms a link with another HTML file or image data located at a network address. In addition, the function to transfer the data located there by specifying the URL is http (Hyper Text).
Provided by the Transfer Protocol). Data can be collected by creating a correspondence table of each data and URL and transferring by http based on this. The correspondence table can be obtained by using a search program or the like for performing manual work or acquisition of an existing URL.

An embodiment of the present invention will be described below according to each step.

FIG. 2 is a flow chart for explaining each process of the image information input section 1. Steps (hereinafter, simply referred to as “s”) 201 to 207 are processes for collecting image information from the network.

First, each image data image [i]
{I: 1 ,. . . , N} image number i and the network addresses add1 [i] and add1 [i] of the image.
The correspondence table as shown in FIG. 12 of the network address add [2] of the HTML file in which is written is previously stored in the network address memory 2 (s20
1). Next, after initialization to i = 1 (s202), the data transfer unit 2 transfers image data from add1 [i] to this apparatus by http (s203) and stores the image data in the image data memory 4. (S204) Save.

Next, the data transfer unit 2 transfers the HTML file from add2 [i] by http to this apparatus (s205), and additionally stores the image auxiliary information in the image auxiliary information memory 3 (s206). The image data stored in the image data memory 4 is the number of pixels in the vertical and horizontal directions and each element of red (R) green (G) blue (B) is 8 bits (256 gradations from 0 to 255). 13 is a full-color image represented by digital data, and the image auxiliary information is the URL of the image data, the URL of the HTML file in which it is described, and the title described in the HTML file, and is shown in FIG. Is stored as

Image number i is compared with N (s207), and if different, i is incremented by 1 and image information input is continued (s208),
If they are equal, the image information input is terminated and the process proceeds to image feature extraction. FIG. 3 is a flowchart for explaining each process of the image database construction unit 4. s331-309
Is a process for extracting image features from image data.

First, in the image feature extraction unit 8, i = 1
After initialization (s301), image [i] is read from the image data memory 6 (s302). Then j =
After initializing to 1 (s303), the program F [j] {j = 1 ,. . . , M} from the image feature extraction method description memory 6 (s30
4) Then, the feature is extracted from the image [i] by executing it, and then additionally stored in the image feature amount memory 7 (s305).

A program F [j] describing the image feature extraction method stored in the image feature extraction method description memory 6
{J = 1 ,. . . , M} are programs for performing area extraction, aspect ratio extraction, saturation extraction, brightness extraction, edge extraction, and representative color calculation as shown in FIG. 15, for example. For the image area and the image aspect ratio, the number of pixels in the horizontal and vertical directions is read from the image data memory 6 and the product and quotient thereof are calculated.
The saturation extraction, the brightness extraction, and the edge extraction read the RGB values of all the pixels from the image data memory 6, and calculate the average of the saturation, the brightness, and the number of edge pixels of all the pixels from 0.0 to 1.
A value in the range of 0 is calculated. Representative color extraction is RG of all pixels
The B value is read, a histogram relating to color is created, and the most frequent color and its frequency are calculated. At this time, the frequency is calculated as a value in the range of 0.0 to 1.0 where the total number of pixels is 1.
For example, as shown in FIG. 13, 64 pixels in the horizontal direction and 32 pixels in the vertical direction.
Image feature values obtained from image data of pixels (for example, area 2048, aspect ratio 2.0, saturation 0.8, brightness 0.4,
Edge 0.3, representative color red, and its occupation ratio 0.6
Etc.) are stored in the image feature memory 7.

After the image feature amount is stored in the image feature amount memory 7, the image feature extraction method number j is compared with m (s30
6) If different, j is incremented by 1 and image feature extraction of image [i] is continued (s307). If they are equal, image is extracted.
The image feature extraction of [i] is completed, and the image numbers i and N are compared (s308). If different, i is incremented by 1 and the image feature extraction is performed on the next image (s309). If they are equal, the image feature extraction is terminated and the process moves to image category detection.

FIG. 4 is a flow chart for explaining each process of the image category detecting section 11. s401-40
Reference numeral 9 is a process for detecting an image category from the image feature.

In the image category detection unit 11, i = 1
After initialization (s401), the image feature of image [i] is read from the image feature memory 7 (s402). Next, after initializing to j = 1 (s403), the image category detecting method C [j] {j = 1 ,. . . , N} is read from the image category detection method description memory 12 (s404),
From the image [i] feature quantity, an image category such as C
After detecting the semantic labeling of G, photos, icons, etc., it is additionally stored in the image category memory 10 (s4
05).

Image category detection method C [j] {j = stored in the category detection method description memory 12
1,. . . , N} is a program that determines the value range of each image feature amount corresponding to each category and whether or not the feature amount of the target image exists in the value range. For example,
For the feature amount of the image feature amount memory of FIG. 13, a comparison is made with the range of each image feature amount shown in the correspondence table of FIG. 14 for each category, and CG and an icon are obtained as a category, and FIG. Is stored in the image category memory as follows.

Further, when storing the image data of each image obtained by the above processing, the feature amount, the category and the auxiliary information in each memory, the address is stored in the index memory as shown in FIG. It is used when processing such as search and display described later.

After the image category is stored in the image category memory 10, the image category detection method number j is compared with n (s
406), if different, j is incremented by 1 and image category detection of image [i] is continued (s407), and if equal, im
The image category detection of age [i] is completed, and the image number i
And N are compared (s408). If different, i is incremented by 1 and the image category is detected for the next image (s40
9), if they are equal, the image category extraction is ended, and the image database construction is ended (s410).

FIG. 5 is a flow chart for explaining the processing operation of the image retrieving section 14, and s501 to s507 are the processing for setting the retrieval range for each data.

First, the user input unit 25 selects whether or not to perform image feature search (s501). When performing the image feature search, the image feature search setting unit 15 sets the range of the feature amount to be searched (s502). Next, in the user input unit 26, it is selected whether or not the image category search is performed (s503). When performing an image category search, the category of the category to be searched is set in the image category search setting unit 19 (s504). Next, the user input unit 27 selects whether or not to perform image auxiliary information search (s505). When performing the image auxiliary information search, the image auxiliary information search setting unit 20 sets a search request (s506). The determination order of each search does not affect the search method, and the above is an example. The search method and its search contents are additionally stored in the used search method description memory 17 (s507), and the process proceeds to the search process.

FIG. 6 is a flow chart for explaining the processing operation of the matching image search unit 15, and s601 to s609 are search processing for each data.

In the matching image retrieval unit 15, first, i = 1
After initialization (s601), the image category of the image image [i] is read from the image category memory 10 (s602), the image feature is read from the image feature memory 7 (s603), and the image auxiliary information is loaded into the image auxiliary information memory. It is read from 10 (s604). It is determined whether the read data matches the search condition (s605). If the read data matches, the image data image [i] is read from the image data memory 6 and additionally stored in the matched image memory 16 (s606). In the reading process of s602 to 604, items not set in s501 to s506 can be omitted. Next, the image numbers i and N are compared (s607). If they are different, i is incremented by 1 and the next image is searched (s608). If they are equal, the search is terminated and the matching image in the matching image memory 1 is matched. The image is displayed on the image display unit 29.
Is displayed on the screen (s609).

FIG. 7 is a flow chart for explaining the processing operation of the image selecting section 30, where s701 and 702 are search matching image displays, s703 is image selection, s704 is similarity search selection, and s705 to 707 are image information displays. Step s708 is a process continuation selection process.

First, M matching images image [i] {i = 1 ,. . . , M} is read (s
701), and the image is displayed on the image display unit 29 (s702).

Next, the image selecting section 30 selects the image image [t] from the displayed M images. The selection method may be, for example, determination based on the degree of conformity or arbitrary selection by the user. Next, the process selection unit 32 selects a process for the image [t].
If similar image search is selected (s704), the process moves to the similar image search setting unit 18. If you select the image information display (s
705), the image information link unit 22 reads the image information of the image [t] from the address of the index memory 13 (s706), and the image information display unit 28 displays the image information.
Is displayed (s707). If the selection process is continued, the process returns to s703, and if finished, the process for the search result is finished (s708).

Next, each processing of the similar image searching section 21 will be described based on the flowchart of FIG.

First, the similarity measuring method and its range are set (s801). Next, the compatible image information memory is initialized (s802), and after initialization to i = 1 (s803), i
Image information relating to the similarity measurement of the image [i] is read (s804). It is determined whether or not the search conditions are met (s805), and if they match, the image information of the image [i] is stored in the compatible image information memory (s806). Next, the image numbers i and N are compared (s807). If they are different, i is incremented by 1 and the next image is searched (s8).
08), if they are equal, the search is ended and the image is displayed on the image display unit 29 (s809).

The user interface section 23 has a GUI (Graphical User) as shown in FIG.
It is also possible to improve the operability of the user by displaying each search method as an icon on the menu screen by the interface) and selecting and inputting it together with the value range. FIG. 17 shows that an item surrounded by a square is selected, and the saturation is 0.2 to 0.
5. Image feature search by the range of lightness 100 to 400 and category search by the icon category are selected.
This search content is stored in the search method description memory as shown in FIG. FIG. 18 is an example in which the image display unit 29, the image selection unit 30, and the image information display unit 31 are displayed on a GUI menu screen. FIG. 18 shows an example in which the rightmost image is selected from the four images displayed in the upper half of the screen and the information of the selected image is displayed in the lower half of the screen.

FIG. 19 shows an example in which the image display unit 29, the image selection unit 30, and the similar image search setting unit 21 are displayed on the menu screen by the GUI. In FIG. 18, the rightmost image is selected from the four images displayed in the upper half of the screen,
It is an example in which search content settings for similar images are displayed in the lower half of the screen.

FIG. 9 is a flow chart for explaining the processing operation of the image category detection method registration section 35.

First, the used search method is read from the used search method description memory 14 (s901), and the search method to be registered is selected (s902). Examples of the selection of the search method include, for example, determination of suitability of search results in addition to the user's arbitrary selection. In the user input unit 24, an image category name is given to the selected search method (s9
03). The image category detection method registration unit 35 registers the image category name and the search method in the image category detection method description memory 9 (s904). If another search method is to be registered, the procedure returns to s903, and if not registered, the registration operation is terminated. For example, in FIG. 19, the search content stored in the search number 1 is displayed from the usage search method description memory 14, and the user inputs the category name icon-jimi1. After inputting, by pressing the register button, the category name and its search content are stored in the image category detection method description memory.

FIG. 10 is a flow chart for explaining the processing operation of the image feature extraction method registration unit 34.

The method existing in the image feature extraction method description memory 6 is combined or a newly created program, for example, describing a method for extracting a texture from an image is prepared in advance (s1001).

Next, in the user input unit 28 as shown in FIG. 20, the network address of the program and the image feature amount name are input (s1002), and the image feature extraction method registration unit 3 is entered.
In step 4, the program is transferred to this apparatus by http and registered in the image feature extraction method description memory 6 as shown in FIG. 15 (s1003). If another extraction method is to be registered, the process returns to step s1001, and if not registered, the registration operation ends.

[0059]

According to the present invention, image data on a network and auxiliary information accompanying it are collected, and an image feature can be obtained from the image data and an image category can be obtained from the image feature. This makes it possible to automatically obtain an image database from scattered image data on the network without enormous time and labor. Furthermore, the user can obtain desired image data by setting the search value for the image feature, the image category, and the image auxiliary information, performing the search, and displaying the search result.

Further, a method of detecting the image category is stored,
By reading out, it is possible to use a plurality of image category detecting methods while additionally registering or changing, and it is possible to flexibly expand the apparatus.

Further, the user can register the image category detecting method. As a result, each user can be expanded as a search device that is easy to use. Further, by registering, storing, and reading the used search method, the history of the search method can be referred to, and the input work of the user can be reduced.

An image category name is added to the search method, and the image category name and the search method are registered. As a result, a search method showing a desired result can be newly registered as an image category, and the performance of the search device can be improved.

Further, by storing and reading out the image feature extraction method of each image feature, a plurality of image feature extraction methods can be used while performing additional registration or change, and a flexible expansion of the apparatus is possible. Can be done.

Further, the image feature extraction method can be registered. As a result, each user can be expanded as a search device that is easy to use.

By associating the retrieved image data with the image information and displaying the image information, it is possible to retrieve the image and simultaneously obtain the associated image information. As a result, the user can obtain the address on the network, the title name, etc. through the image. Further, the similarity between the searched image data and each image data is obtained, and the similar image is obtained by the similarity search means. As a result, a search similar to the intuition of the user can be performed.

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating an embodiment of an image search device according to the present invention.

FIG. 2 is a flowchart illustrating a processing operation of the image information input unit 1 of the present invention.

FIG. 3 is a flowchart illustrating a processing operation of an image database construction unit 4 of the present invention.

FIG. 4 is a flowchart illustrating a processing operation of the image category detection unit 11 of the present invention.

FIG. 5 is a flowchart illustrating a processing operation of the image search unit 14 of the present invention.

FIG. 6 is a flowchart illustrating a processing operation of a matching image search unit 15 of the present invention.

FIG. 7 is a flowchart illustrating a processing operation of the image selection unit 30 of the present invention.

FIG. 8 is a flowchart illustrating a processing operation of a similar image search unit 21 of the present invention.

FIG. 9 is a flowchart illustrating a processing operation of an image category detection method registration unit 35 of the present invention.

FIG. 10 is a flowchart illustrating a processing operation of an image extraction method registration unit 34 of the present invention.

FIG. 11 is a diagram illustrating a network environment using World Wide Web as an example.

FIG. 12 is a diagram illustrating contents stored in a network address memory 2 of the image information input unit 1 of the present invention. .

FIG. 13 is a diagram illustrating the contents stored in the index memory 13 obtained by the image database construction unit 4 of the present invention and each memory associated therewith.

FIG. 14 is a diagram illustrating a correspondence relationship between an image category of the present invention and an image feature amount used for detecting the image category.

FIG. 15 is a diagram for explaining the contents held in the image feature extraction method description memory 9 of the present invention.

FIG. 16 is a diagram for explaining the contents stored in the usage search method description memory 17 of the present invention.

FIG. 17 is a diagram showing an example of a search setting screen of the present invention.

FIG. 18 is a diagram showing an example of a search result display screen and an auxiliary information display screen of the present invention.

FIG. 19 is a diagram showing an example of a search result display screen and a similar image search setting screen of the present invention.

FIG. 20 is a diagram showing an example of an image category registration screen of the present invention.

FIG. 21 is a diagram showing an example of an image feature extraction method registration screen of the present invention.

[Explanation of symbols]

 1 image information input unit 2 network address memory 4 image database construction unit 9 image feature extraction method description memory 11 image category detection unit 13 index memory 14 image search unit 15 matching image search unit 17 usage search method description memory 21 similar image search unit 23 User interface unit 30 Image selection unit 33 Classification method registration unit 34 Image extraction method registration unit 35 Image category detection method registration unit

Claims (9)

[Claims] 1. An apparatus for retrieving image data on a network, image information input means for collecting image data on the network and auxiliary information accompanying the image data, image data storage means for storing the image data, and the image. Image auxiliary information storage means for storing auxiliary information, image feature extraction means for extracting image features from the image data stored in the image data storage means, and image feature obtained by the image feature extraction means Image feature storage means, image category acquisition means for obtaining an image category from the image features stored in the image feature storage means, image category storage means for storing the image category obtained by the image category acquisition means, Image category reading means for reading the image category stored by the image category storing means; Search image feature setting means for setting search values for image features; search image category setting means for setting search values for the image categories; search image auxiliary information setting means for setting search values for the image auxiliary information; A search means for performing a search based on a search value for the set image feature, a search value for the image category, or a search value for the image auxiliary information; and a display means for displaying a search result by the search means. An image retrieval device characterized by. 2. The image category search method storage means for storing a plurality of search methods for each image category, wherein a plurality of image data are searched based on the stored search method. Image retrieval device. 3. The image search device according to claim 1, further comprising image category search method registration means for registering an image category search method, wherein the image category search method can be arbitrarily registered. 4. A used search method registration means for registering a used search method, and a used search method reading means for reading out a used search method registered in the used search method registration means. The image search device according to claim 1, wherein 5. An image category name registration means for registering an image category name in the search method, and an image category search method registration means for registering the image category obtained by the image category name registration means and the search method therefor. The image retrieval device according to claim 1, 3 or 4, characterized in that. 6. An image feature extraction method storage means for storing an image feature extraction method for each image feature, and an image feature extraction method reading means for reading out each image feature extraction method from the image feature extraction method storage means. Claim 1 characterized by the above-mentioned.
The image search device according to the above. 7. The image search apparatus according to claim 1, further comprising image feature extraction method registration means for registering an image feature extraction method, wherein the image category extraction method can be arbitrarily registered. 8. The image according to claim 1, further comprising image information linking means for associating the retrieved image data with the image information, and displaying image information corresponding to the retrieved image data on the display means. Search device. 9. The apparatus according to claim 1, further comprising a similarity measurer for the searched image data and each image data, and a similarity searcher for performing similarity search based on the similarity obtained by the similarity measurer. The image search device according to items 1, 5 and 7.