JP2001005822A - Image display method and apparatus and search service using the same - Google Patents

Abstract

(57) [Summary] [Problem] To display a large number of result images of an image search by an arbitrary search method in a list so that a searcher can easily grasp the images. Kind Code: A1 A keyword is previously associated with a feature amount by statistical processing, and in a search image obtained by an arbitrary image search, a feature amount of a keyword that effectively represents a difference between individual images is dynamically determined for each search. Automatically selected and used for the axis of display. Further, the image is arranged at coordinates according to the feature amount in the display space formed by the axis. Further, the feature amount of a keyword expressing the entire image group of the search result is dynamically and automatically selected for each search, and the keyword is presented to the searcher.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image display method and apparatus in an arbitrary search system that outputs an image group as a search result, and a search service using the same.

[0002]

2. Description of the Related Art In recent years, with the increase in multimedia contents, a technique for accessing an image database using various data such as keywords and image feature amounts and performing an image search has appeared.

[0003] However, since the search is not always the one in which the intention of the searcher is accurately conveyed to the system and reflected in the search result, it is difficult to display a small number of search results to reach the image desired by the searcher. Not necessarily. In some cases, a searcher who has ambiguous search intentions needs to have a process of consolidating search intentions while accessing various images. Under such a situation, it is desired by a searcher to present as many search result images as possible to the searcher so that the searcher can easily reach an image corresponding to the search intention.

For example, Japanese Patent Laid-Open No. 5-26 discloses a method for effectively presenting a list of images of search results to a searcher effectively.
6092 and JP-A-6-96185. These methods change the page turning time and the image size according to the number of search results. However, when simply trying to list many images in the search results, the number of pages increases, or a large number of images are unordered on one page, so that a searcher cannot search for individual images in the search results. And the effect of the list is diminished.

[0005] Therefore, a method is conceivable in which the arrangement of the retrieved images is determined in accordance with the properties of the retrieved images themselves, so that the characteristics or tendency of the retrieved image group can be easily understood.
For example, JP-A-5-282375 and JP-A-8-249352 disclose such measures. In the former, information having two-dimensional attribute information is selected from image feature amounts, a search is performed by designating a range of the selected two-dimensional information, and the feature amount is determined in a space formed on the two axes. An image is arranged and displayed at the coordinate values according to the coordinates. In this idea, the arrangement of the image itself is improved. However, since the image is displayed only by the feature amount having two-dimensional information, the limitation of the feature amount usable for display is very large. The latter performs a search by selecting two dimensions from the physical features directly extracted from the image or the kansei features learned from human sensibility and outputting them in order from the one with the shortest distance. The images are displayed in a list along the axis in a densely arranged order without overlapping. In this method, although the tendency of the feature amount of two axes can be grasped, the quantitative relation of the feature amount between images is ignored. In addition, if many images with similar tendencies are searched because the list is displayed in an orderly manner, similar images occupy a large area of the screen of the list display, and effective display cannot be performed. .

In each method, since the feature amount used for the search is also used for direct display, it is effective for grasping the tendency of the image group of the search result in the database, but it is effective. The display does not always represent the difference between the individual images in the group. In the first place, the images obtained by the search are, of course, images with similar tendencies or common points, so it is difficult to distinguish them by directly using the features used for the search for display. You will place individual images with minor differences.

[0007]

On the other hand, the inventors of the present application have proposed Japanese Patent Application No. 9-336751 as a method for effectively expressing the difference between individual images in an image group of a search result. According to this, by dynamically performing statistical processing on a group of images of a search result for each search, the image is converted into a feature amount that effectively expresses a difference between the search images, and an image is formed on a space created by the axis. Can be arranged and displayed.

[0008] However, the display according to JP-A-5-282375 or JP-A-8-249352 is easy to understand the meaning of the axis at the time of display if a feature amount which is easy for humans to understand is prepared. Japanese Patent Application No. 9-336751
In the method of (1), a feature amount that effectively represents the difference between images is derived by statistical processing inside the computer, and the image is arranged and displayed with this feature amount as an axis, so that the meaning of the axis is not necessarily understood by humans. It is not always easy. Further, in terms of an image search method and a display method of a search result, JP-A-5-282375 and JP-A-8-249352 disclose.
Also, Japanese Patent Application No. 9-336751 is closely related to each other, and therefore, when performing a search, if an advantage of these display methods is to be obtained, the image search method is restricted.

As described above, the prior art has advantages and disadvantages with respect to a method of displaying a large amount of image search results by an arbitrary search method in a list so that a searcher can easily grasp the result. The problems of these prior arts are summarized, and the main conditions for an effective list display method are summarized as follows. (1) Many search results can be presented to a searcher. (2) The quantitative relationship between the feature amounts of the search image group is also reflected in the display. (3) A display method that effectively expresses a difference between individual images in an image group with respect to an image group searched by an arbitrary search method. (4) The axes used for display are easy for humans to understand.

It is an object of the present invention to propose an image display method and apparatus which simultaneously satisfy the above-mentioned problems, and a search service using the same.

[0011]

According to the present invention, a keyword or a symbol is previously associated with an image feature amount by statistical processing, and the difference between individual images can be effectively reduced in an image search by an arbitrary image search method. The feature amount of a keyword or symbol to be expressed dynamically is automatically and dynamically selected for each search and used as a display axis, and the keyword or symbol is presented to a searcher. Also, the images of the image group of the search result are arranged at coordinates according to the feature amount in the display space formed by the axis. further,
The keyword or symbol of the coordinates automatically selected and used for the display axis can be selected by the searcher.

[0012]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Preferred embodiments of the present invention will be described below in detail with reference to the accompanying drawings.

FIG. 1 is a schematic diagram showing a graphic user interface (hereinafter abbreviated as GUI) according to an embodiment of the present invention. The figure shows several keywords 1 representing the entire search result image group for an image group output from an arbitrary search system that outputs a search result image group.
08 and some keywords 10 in the keyword group that best represent the differences between the displayed individual images of the image group.
5,107 are displayed on the panel, and the latter keyword group 1
The user selects two keywords from 05 and 107,
Display space 106 spanned by the keyword features
Above, the image group is represented as a scatter diagram.

The display space 106 is, for example, a part of a display device as an output device of a computer, and is a portion for displaying a group of images as a scatter diagram on a two-dimensional plane. Keyword group 1
05 and 107 are keywords to be selected as the X axis and the Y axis when the display space is a two-dimensional space, and are displayed on a part of a display device as an output device of a computer, for example. Area functions as a button, and one keyword is selected by clicking or the like. In the figure, a state in which “SEA” is selected on the X axis and “SKY” is selected on the Y axis is shown by enclosing each keyword in a thick line frame. The display space 1
Reference numeral 06 denotes a space spanned by the feature amount vectors of the respective keywords. Each of the plurality of images represented by reference numeral 102 predicts whether or not a keyword is assigned to each image from the image feature amounts. Is arranged at a position corresponding to the magnitude of the predicted value. Keyword group 10
Arrow buttons 111-114 for selecting a keyword to be displayed are displayed on both sides of 5, 107, and are used to express the difference between the displayed images in the image group by clicking these buttons. Keywords can be changed. The change by selecting the keywords on the X axis and the Y axis only changes the position where each image is arranged.

On the right side of the panel, a keyword group 108 representing the entire retrieved image group is shown. Keywords that are predicted to explain the image group more appropriately are displayed at the top.

The scale of the display space 106 can be changed by the scroll bar 101.
This can be used for grasping the entire distribution of the image group or for enlarging a part of the image for detailed viewing. Also, arrow button 1
With 03, the display area of the display space can be moved. That is, for example, when the right arrow is clicked, the area outside the right of the display area moves toward the center. "CENTE
When the R "button 104 is clicked, it is reset so that the center of the display space coincides with the center of the display area 106.

With the functions of the above-described GUI, the searcher can change the displayed image group as a search result by changing the keyword for display or the keyword group, moving the display space, and enlarging / reducing. And make it possible to reach more satisfactory search results.

FIG. 2 is a schematic diagram showing a data flow according to the embodiment of the present invention. 202, 203, 204 and 20
Reference numeral 5 denotes a keyword database, a sentence database, a voice database, and an image feature amount database. Reference numeral 227 denotes an image database in which images displayed in the display space 106 to be displayed according to the search result are stored as pairs of image data and image data names or image numbers. The databases 201 to 205 store image data names or image numbers of the image database 227 in pairs. For example, the image feature amount database 205 will be described in association with an example of selection and display of a keyword based on an image feature amount described later. The vertical axis indicates an image data name or an image number, and the horizontal axis indicates a feature of a mathematical expression for calculation described later. This is a database in which the quantity vectors x _ik are arrayed. Reference numeral 217 denotes a keyword feature amount database. This database characterizes the present invention, and is configured independently of the contents of the image database 227 described above. For example, it is created in advance from an image and a keyword attached to the image, or a data group having simple sentences explaining the image. If there is a simple sentence explaining the image, it can be decomposed into appropriate words by morphological analysis, and can be used as data on the relationship between the image and the keyword. On the other hand, several methods are known for analyzing an image and evaluating it as an image feature amount. Therefore, these methods are used in advance so that the presence or absence of an appropriate keyword corresponding to the image can be predicted using the image feature amount as a medium. Statistical processing can be performed. An appropriate number of pairs of keywords and image feature values that have been statistically processed in this way are stored. The feature amount of the keyword feature amount database 217 is calculated using feature amount vectors b ₀ , b _k, etc.
Used as data about The calculation represented by (Equation 2) can be said to be an example of statistical processing of a keyword.

When a user issues a search request 201 using a keyword, a sentence, a voice, an image, or the like as a query, data is read from a corresponding database 202-205 by a search method 206 corresponding to the search request to perform a search.
Reference numerals 207, 208, 209, and 210 denote the flow of keywords, sentences, voices, images, or feature amounts that have been converted and stored when searching. In the present invention, since the search method may use whatever the user wants to use, 206 is only a variety of image search systems and will not be described in detail. Of course, this does not mean that a database of keywords, sentences, and sounds must be provided. The search result is output in the form of a data name of an image in the image database 227 or a number capable of specifying an image in the image database. Each search result is a keyword, a sentence, a voice, and an image. Each data in the database corresponding to the former three and the image in the image feature amount database 205 are previously associated with each other. In addition, an image can be displayed by associating a search result with a keyword based on the image feature amount database 205.

The search result 211a output in the form of an image data name or an image number is sent to a function 213 for calculating an average value, a variance value, and a covariance value of a plurality of feature amounts for the image group of the search result. . Further, to the function 213, data relating to the image feature x corresponding to the image data name or the image number is sent from the image feature database 205 as indicated by 212. From the function 213, average value data 214 and variance-covariance value data 215 for the image group of the search result are obtained.

The keyword feature database 217 is
As described above, the image feature amount of each keyword that has been statistically processed in advance is stored so that the image feature amount and the presence or absence of an appropriate keyword corresponding to the image can be predicted. As shown in 216, the data on b in each keyword in the keyword feature amount database 217 is sent to the function 218 for calculating inner products and the like together with the average value data 214 and the variance-covariance value data 215 for the image group of the search result. As an operation result by the function 218, an average value 219 of keyword prediction values for the image group of the search result,
The variance 220 is output with the corresponding keyword.

Reference numeral 221 denotes a selection function, and a designated number is selected from those having a large calculation result of the average value 219 of the keyword prediction values by the function 218 and the average value of the variance values 220. The keywords selected based on the average value 219 of the keyword prediction values are transferred to the designated number (S in FIG. 3) representative keyword presenting function 224 as shown at 222. The number of keywords selected based on the variance value 220 is a specified number (T in FIG.
) Is transferred to the search result display function 225. The representative keyword presenting function 224 calculates the average value 21 of the keyword prediction value.
Since the keyword having 9 is displayed on the GUI as the keyword 108 representing the image group of the search result, the user can grasp as the semantic information representing the image group of the search result (228). These are keywords (223) selected based on the variance value, and the search result display function 225 converts the keywords selected based on the variance value 220 as keywords 105 and 107 for grasping differences in the image group. Display on the GUI. Further, the search result display function 225 includes a feature amount vector x _ik transferred from the search result 211 b and the image feature amount database 205 as indicated by 233, and a selected keyword 2.
And, in response thereto, the data on the keyword features b ₀ , b _k, etc. b transferred as indicated by 231,
The X-axis and Y-axis coordinate values are calculated for each of the images of the search result 211b, and the images are arranged. The user can see the difference between the searched image groups as a scatter diagram displayed on the semantic axis (229). Note that the search results 211a and 211b have the same contents, but are different from each other for convenience of explanation. The sign was used.

FIG. 3 is a PAD diagram showing a processing procedure according to the embodiment of the present invention. The processing in one search and the display of the relation between the searched images is described.

First, a search request is received from the user (3
01), an image search is performed by an arbitrary method, and the image data name is output (302).

Next, an image feature vector corresponding to the image data name is obtained from the image feature database 205 (309). The average value and the variance-covariance matrix of each element of the feature amount vector are updated from the feature amount vector corresponding to the one search result (310). By repeating the processing of 309 and 310 by the number of images of the search result (303), the average value 214 and the variance-covariance matrix 2 in the search result are obtained.
Get 15.

Further, the keyword feature amount database 21
7 to extract one keyword feature vector (31)
1). The average value 219 of the keyword prediction value is calculated from the average of the respective elements of the image feature vector (312).
It compares with the S minimum values stored at the present time, and if it is larger than that, discards the minimum value and stores the average value of the keyword calculated just now. As a result, S keywords are stored from the current maximum value (313). The variance of the keyword prediction value is also calculated from the variance and covariance of each element of the image feature vector (3
14). Similarly to 313, the variance values 220 for the T keywords from the maximum value are stored. These processes are repeated by the number of keywords (304). When this is completed, for all the keywords, S keywords are selected from those having a large average keyword prediction value, and T keywords are selected from those having a large dispersion value.
You can get

Then, a keyword based on the average of the keyword prediction values is presented to the user (305). This becomes the keyword 108 representing the image group of the search result.

Also, a maximum of U keywords are presented to the user from the beginning as a selectable interface based on the variance of the predicted keyword value (306). Next, whether or not the keyword presented to the user has been changed (308) is checked.
Is designated as the initial value on the X axis and the second on the Y axis (3
16). Then, the image feature vector of the search result is projected and displayed as a scatter diagram on the space spanned by the designated keyword feature vector (317). By specifying the initial value (316), some display can be performed immediately after the search without any special operation of the user, and the next operation of the user is prompted. Next, a standby state for obtaining an instruction from the user is entered (318).

When an event is received through the GUI, the event is first determined at 319. If it is a change of the displayed keyword button 105 or 107, the pressed keyword is specified for the axis of the pressed button (320). If it is one of the keyword shift buttons 111-114, the keyword to be presented to the user on the axis of the designated shift button according to the direction of the arrow of the button is the next or next maximum U.
Change to Otherwise, the first and last maximum U respectively
(321). If it is a move button or a scroll bar of the display space, the display space is moved or the display area is enlarged or reduced according to the buttons (322).

These are repeated until the next search (30
7) By doing so, the user can obtain a display from various viewpoints for the same search result, and can use the search result effectively.

The keyword based on the variance in step 306 is a keyword that is presumed to more reflect the difference in the image group of the search result. By arranging the search result in a space spanned by the keyword feature amount,
It is expected that different images corresponding to the presence or absence of the selected keyword are classified and displayed.

The above process is a process for one search, and is executed each time the search is repeated.

Here, steps 312 and 3 in FIG.
The operation related to the process of No. 14 will be described. As an embodiment of the present invention, multiple regression analysis is used as a method for predicting a keyword from an image feature amount. The prediction equation in that case is as shown in (Equation 1).

[0034]

(Equation 1)

Since (Equation 1) is linear, it may not be predictable depending on the keyword. However, it is possible to perform processing so that only the predictable keyword is adopted in advance. Here, it is important to obtain the prediction coefficient b _k for each keyword in advance. It is determined by (Equation 2) in multiple regression analysis.

[0036]

(Equation 2)

When all the elements of the image feature vector are used, if the elements do not contribute to the prediction of the keyword, the prediction performance deteriorates. Therefore, there is a method of selecting elements of the feature amount vector so as to increase the prediction performance. This is a method of selecting an element by F test or information criterion AIC. A specific method is described in, for example, “Multivariate Analysis Handbook” of Hyundai Mathematics Co., Ltd. do not do. In any case, the prediction coefficient b _k =
If it is set to 0, the number of elements of the image feature amount vector adopted for each keyword does not change, so
Easy to handle.

To determine a keyword representing a group of images as a search result, the following equation (3) is used. The average of the predicted keyword values in the search result is obtained by (Equation 3).

[0039]

(Equation 3)

Here, the dash attached to the variable of the average value indicates that all data n when the number of data to be averaged is (Equation 2)
On the other hand, in (Equation 3), since the number of search results is m, they are used to distinguish them.

In order to determine a keyword that more reflects the difference between the image groups of the search result, the equation (Equation 4) is used.

[0042]

(Equation 4)

It can be seen that (Equation 4) requires a larger amount of calculation than (Equation 3). If it is worrisome that the amount of calculation is large due to the capability of the computer or the number of elements of the feature quantity vector, a method in which the value of (Equation 3) is close to 1/2 is adopted as an approximate method. There is a way. If a certain variable takes a binary value of 0 or 1, the variance is P (1-P), where P is the frequency of appearance. This is maximum when P = 1/2, so to select the one with the maximum variance, P = 1/2
Good thing to choose. This method is effective when keyword prediction is successful for all search results. Moreover, the value already calculated in (Equation 3) can be used, and it is not necessary to newly calculate the value, which is very useful for reducing the amount of calculation.

Although an example in which multiple regression analysis is used as statistical processing has been described here, use of secondary discriminant analysis or learning of neural networks of three or more layers as statistical processing is prevented. Not something. Since these perform nonlinear processing, there is a possibility that keywords that could not be predicted when multiple regression analysis that performs prediction by linear processing is used can be predicted.

FIG. 4 is a schematic diagram showing the respective configurations and linkages between the client side and the server side as an embodiment when the display according to the present invention is performed using a network system. Image search client system 4
07 and the image search server system 408 are connected to the network 4
02 is connected. The device 401 is similar to that of FIG.
This is a display and input device that implements a GUI or the like as described in. Input such as selection of a display keyword and movement of a display area is performed on the GUI. It also has an input function to accept search requests for various image searches. Device 40
6 is a network server 403 and a network 402
Is a network client that communicates via.
A user's search request is transmitted to the device 403, a screen is configured according to the consulted image data and keywords, coordinate information of each image, an image arrangement is changed according to information from the device 401, and a display area is moved. Or have the function of The device 403 is a network server that transmits and receives data via the network 402. More specifically, it has a function of receiving a search request here, transmitting a searched image, coordinate information of each coordinate for display, a keyword, and the like. Image database 4
05 stores the image itself, the feature amount of each image previously extracted from the image, and tag information (keywords, sentences, sounds, etc.) for performing various searches. Used for As shown in FIG. 2, the database can be stored separately for each search method or the like, but the entire apparatus can be stored in one database. The device 404 is an image search device that performs a search according to a search request. The image database 405 is accessed, and the search result is passed to the network server 403. In addition to the search function, selection of a representative keyword or a keyword expressing a difference in the image group is also performed here.

FIG. 5 is an NS chart for explaining the communication procedure of the client / server system for displaying the relationship between search result images.

First, a server system (hereinafter, referred to as an S system) 521 is started. The S system is initialized (522), and preparations such as reading a setting file are made. Next, it is in a state of waiting for communication from the client system (523).

Then, the client system (hereinafter referred to as C
When the (system) is started (501), initialization is performed in the same manner as in the S system (502). In communication connection processing 503, a communication connection request is transmitted to the S system that is waiting for communication. The S system establishes a communication connection with the C system that has requested the communication connection in the communication connection process (523). In this case, communication may be waited until there is a connection request, exchange of information with the C system, transmission of a program for a GUI, and the like may be performed. C
In the communication connection processing 503 of the system, the received information is further set and a GUI program is executed. Thus, preparation for image search is completed. From here on, the GUI starts to be activated in the C system, and the S system responds to the request of the C system and executes the processing.

First, in the processing of the C system, a search request for executing a specific search in the presence of various image searches from a user is received (504). This search request is transmitted to the S system (505). This data is
And When the communication waiting S system receives the communication,
First, the received content is determined (524). Communication content is C
If it is 1, the process goes to the "Y" branch to execute an image search (526). In the case of the "N" branch, the process returns to the communication waiting 524 (527). At this time, the error information may be transmitted to the C system, or without transmitting the error information, the time may be returned to the communication connection process 503 by performing a timeout process on the C system side.

After performing image search by going to the "Y" branch in the S system, calculation and selection of representative / display keywords, and coordinates of the search result image are calculated for each of the selected display keywords (528). . Next, the search result image itself is transmitted (529).

When the image of the search result is received in the communication standby process 506 of the C system, a request for a representative / display keyword (C2) is transmitted (507). Then, the C system is in a communication standby state of the representative / display keyword (50).
8). When the representative / display keywords are received from the S system, a request (C3) for the coordinates of each image in each display keyword is transmitted (509), and the communication waits (51).
0). Further, upon receiving the data (510),
Transmit data reception confirmation (C4) to S system (51)
1). Then, a communication disconnection process is started (512).

In the S system, when a communication is received in the communication standby (530) state, the content is determined. If the content is C2, the representative / display keyword is transmitted. If the communication is C3, the coordinates of each search image in each display keyword are transmitted. Is transmitted, and after each end, the process returns to the communication waiting (530) state. If it is C4, the process directly proceeds to the communication disconnection process (535), and the one-time search is terminated, and the process returns to the first communication standby (523). This is repeated until the system ends (536). If it is time to end the system, end processing is performed (537).

In the C system, the representative / display keywords are presented to the user, the keyword designation of the display axis is accepted (513), and the image is arranged and displayed using the coordinates of each search image corresponding to the keyword axis. Is performed (514).
In processing 514, processing such as movement of the display area is also performed. When the display is completed, the process returns to the initial communication connection process (503). This is repeated until the search is completed (515). When the search ends, an end process is performed (516).

In this embodiment, the two-dimensional display surface has
Search results are displayed as the axis display space,
The display space can be uniaxial or triaxial. Moreover, it is easy for the searcher to arbitrarily select this.

[0055]

According to the present invention, an output result of an arbitrary search system that outputs an image group of a search result can be displayed on a two-dimensional surface by using a keyword or a symbol, for which a user can easily understand the meaning. , Advertising and broadcasting industries that frequently use designs because they can quickly list search results and easily prepare for the next search request, or content services industries that sell designs. Not only can be used effectively, but also can be used for a creation support application such as creating a homepage or other designs on an individual basis that runs on a PC.

[Brief description of the drawings]

FIG. 1 is a schematic diagram showing a graphic user interface according to an embodiment of the present invention.

FIG. 2 is a schematic diagram showing a data flow according to the embodiment of the present invention.

FIG. 3 is a PAD showing a processing procedure according to an embodiment of the present invention.
FIG.

FIG. 4 is a schematic diagram showing respective configurations and cooperative relations between a client side and a server side as an embodiment in a case where display of a search result image relationship according to the present invention is performed using a network system.

FIG. 5 is an explanatory diagram of a communication procedure of a client-server system for displaying a relationship between search images.

[Explanation of symbols]

101: scroll bar for changing the scale of the display space, 102: searched image, 103: button for moving the display area of the display space, 104: button for returning to the center of the display area of the display space, 105: display Button for selecting a keyword to be used on the Y-axis; 106: a part for displaying the display space; 107: a button for selecting a keyword to be used for the X-axis of display; 108: a part for displaying a group of keywords representing the image group; : Search request from user, 202: keyword database, 203: sentence database, 204: audio database, 205: image feature amount database, 206: various image search methods, 207: flow of keywords or feature amounts created from keywords , 208: Flow of a sentence or a feature generated from the sentence, 20 : Feature quantity of flow made from speech or voice, 210: image feature amount of the flow, 211:
Image data name of search result or number specifying image,
212: Flow of feature amount of image corresponding to search result, 21
3: Average of each element of the feature vector in the search result;
Calculation of variance / covariance, 214: Flow of average value of each element of feature quantity vector, 215: Flow of variance / covariance of each element of feature quantity vector, 216: Flow of image feature quantity for each keyword, 217: Keyword feature Quantity database, 2
18: Calculation of average value and variance value of each keyword predicted value, 2
19: average value in the search result of each keyword prediction value,
220: variance value in the search result of each keyword predicted value, 221: a process of selecting a specified number from the maximum value based on the average value and the variance value, 222: a keyword selected based on the average value, 223: a variance value Keyword and feature quantity selected based on it, 224: Presentation of representative keyword, 22
5: GUI for display of display keywords and display of search results,
226: Flow of the search result image itself, 227: Image database, 228: Semantic information of the entire image group of the search result,
229: semantic information indicating a difference in the image group of the search result, 2
33: Feature vector x _ik to be transferred, 301: _Acceptance of search request from user, 302: Image search and output of data name of search result, 303: Repeat by the number of search results, 3
04: repeat by the number of keywords, 305: present keywords based on the average value to S users from the maximum value, 306: present keywords based on the variance value to the T users from the maximum value, 3
07: Instruction of display keyword from user, 308: Display of image of search result, 309: Acquisition of image feature vector, 310: Total of average and variance / covariance of each element of image feature vector, 311: Keyword feature Acquisition of quantity vector, 312: Calculation of average of keyword predicted value, 313:
Save S keywords from the maximum value, 314: Calculate variance of predicted keyword values, 315: Save T keywords from the maximum value.

 ──────────────────────────────────────────────────続 き Continued on the front page F term (reference) 5B075 ND08 NK06 NR06 NR20 PP02 PP03 PP12 PP22 PP28 PQ02 PQ16 PQ40 PQ46 PQ48 PQ60 5E501 AA01 AC15 AC33 BA03 FA14 FA23 FA44 FA48 FB04

Claims (8)

[Claims] 1. A method for displaying an image of an image group of a search result obtained as a result of a search from an image database by an arbitrary method, wherein a feature amount for the image group of the entire image database and a keyword that can be assigned to the image Alternatively, a database in which the correspondence between the symbol and the image feature amount is associated in advance by statistical processing is provided, and the feature amount of a keyword or a symbol expressing the difference between the feature amounts of a plurality of images in the search result is extracted between the feature amounts. In the order sensitive to the difference, only the maximum number specified from the top is dynamically automatically selected for each search, the selected keyword or symbol is presented to a searcher, and the presented keyword or symbol is presented. Display space consisting of keywords or symbols selected by the searcher from Search image display method characterized by arranging displayed as a scatter diagram reduced images of a plurality of images of the search results. 2. When it is determined that there is no keyword or symbol feature suitable for expressing a difference in the feature of the searched image group, the keyword or symbol feature automatically selected is used instead of the keyword or symbol feature. A statistical process is performed dynamically on the entire feature amount of a plurality of searched images for each search, and only one or a plurality of feature amounts having a large contribution from a multidimensional space calculated therefrom are used. Item 1. The search image display method according to Item 1. 3. The search image display method according to claim 1, wherein the presented keyword or symbol can be changed by a searcher. 4. A feature amount of a keyword or a symbol representing the entire feature amount of a plurality of searched images is automatically selected for each search dynamically by a maximum number specified from a high order, and the keyword or the symbol is selected. The search image display method according to any one of claims 1 to 3, which is presented to a searcher. 5. A display device for displaying an image of an image group of a search result obtained as a result of a search from an image database by an arbitrary method, wherein a feature amount for the image group of the entire image database and a keyword which can be assigned to the image Alternatively, a database in which the correspondence between the symbol and the image feature amount is associated in advance by statistical processing is provided, and the feature amount of a keyword or a symbol expressing the difference between the feature amounts of a plurality of images in the search result is extracted between the feature amounts. Means for automatically selecting only the maximum number specified from the top in the order sensitive to the difference for each search, means for presenting the selected keyword or symbol to a searcher, the keyword or symbol presented Display space consisting of keywords or symbols selected by the searcher from Search image display apparatus, characterized in that it comprises means for positioning and displaying the scatter diagram reduced images of a plurality of images of the search results. 6. A means for automatically selecting a maximum of a specified number of keywords or symbols representing the entire feature amount of a plurality of searched images from a higher order for each search, and the keyword 6. The search image display device according to claim 5, further comprising: means for presenting a symbol to a searcher. 7. The search image display device according to claim 5, further comprising means for changing the presented keyword or symbol by a searcher's operation. 8. An arbitrary search service for outputting an image group of a search result obtained as a result of a search from an image database by an arbitrary method, wherein a feature amount for an image group of the entire image database and a keyword which can be assigned to an image. Alternatively, a database in which the correspondence between the symbol and the image feature amount is associated in advance by statistical processing is provided, and the feature amount of a keyword or a symbol expressing the difference between the feature amounts of a plurality of images in the search result is extracted between the feature amounts. In the order sensitive to the difference, only the maximum number specified from the top is dynamically automatically selected for each search, the selected keyword or symbol is presented to a searcher, and the presented keyword or symbol is presented. From keywords or symbols selected by the searcher from The search service side has a function of arranging and displaying reduced images of a plurality of images of the search result as a scatter diagram on a display space, and the searcher uses this via a user interface and a communication line. Image search service.