JP2001229382A - Information storage device, information retrieval device, information storage method, information retrieval method and recording medium with these methods stored therein - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide the information storage device/method, information retrieval device/method and a recording medium storing these methods which can shorten the time required for calculating similarity of objects and also can present the information on the photographed objects to users. SOLUTION: Plural images obtained by photographing previously an object in plural conditions are inputted and the feature value of images are extracted (steps 11 and 12), and the information on the object and its relative objects are generated and stored in a data base with each feature value relative to the same object (steps 13-15). At the same time, the feature value relative to the images having high similarity about the same object are selected or summarized, When a user inputs the image of a desired object for acquiring its information, the user retrieves the similarity for each feature value stored in the object in response to the feature value acquired from the image. Thus, the information on an object stored in the object that is coincident with the retrieved feature value or the feature value having high similarity is presented to the user.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an information storage device, an information retrieval device, and an information storage device for presenting information on an object to a user based on an image, video, or the like of the object given by the user. The present invention relates to a method, an information search method, and a recording medium on which these methods are recorded.

[0002]

2. Description of the Related Art A dictionary or picture book stored in advance on a CD-ROM or a network is searched based on a search term using an input device such as a keyboard. There are a large number of information search devices that present the information to the user. On the other hand, research and development on similar object search, which uses an image as an input and searches for similar images or information closely related to images, are widely performed in response to search targets that are difficult to express with search terms. I have. In addition, some features are extracted in advance from a plurality of objects using an image input device, stored in a database (DB) in association with information related to the objects, and the features calculated by the user from real-world objects are stored. A method of performing a similarity search on a basis is widely known.

As a first method for performing a similarity search based on features calculated from objects in the real world by a user, a gray-scale image obtained by photographing an object without extracting difficult-to-implement geometric features is converted into a multidimensional vector. There is a method in which the feature amount of a video of the same object continuously photographed while changing the composition of the object in advance is mapped to an eigenvector space to determine a trajectory, and a similarity search is performed based on the trajectory (for example, And a high-speed eigenvector calculation device for a moving image recognition device disclosed in Japanese Patent Application Laid-Open No. 8-329247).

As a second method, in searching for similar objects, a similarity between a search key object extracted from an image and given and an object in the DB is determined from a key object and individual DB objects. There is a method of making a determination based on the degree of similarity between vectors of extracted feature amounts (for example, a similar object search device disclosed in Japanese Patent Application Laid-Open No. H10-154149). At this time, various types of feature amounts used for digitizing the object, such as color tone, brightness, saturation, and the direction of a line in an image, when the object is an image, are used.

[0005]

However, in the first method, the eigenvector is calculated, and the similarity is obtained by expanding the contour in the similarity search. In general, the eigenvector of the multidimensional data is obtained. The calculation requires considerable calculation time. For this reason, this method takes a grayscale image as input, and detects the position and direction of a specific object, derives the similarity between the input image and the limited object, or detects similar movement based on the movement of the object. It is mainly used for the method, and is not suitable for a method of presenting information on a similar object from many candidates having different colors and shapes.

In the second method, an object to be searched is limited, a feature amount extracted from an image obtained by photographing a plurality of objects in advance is stored in a DB, and extracted from an image photographed separately. By retrieving based on the obtained feature amount, a certain degree of object identification can be realized at high speed. However, in general, in the case of an object having a different shape, color, or the like depending on a composition to be photographed, or an object whose shape changes, a feature amount extracted from an image greatly changes depending on a situation. For this reason, even if the images of the same object are captured, the similarity between the feature amounts extracted from the images captured with different compositions or shapes is often extremely low, and the user can select an arbitrary composition or shape. It is not suitable for a method of presenting information on a corresponding object or a similar object using an image of the object.

Further, as in the first method, images of the object are continuously photographed in various compositions and shapes while changing the composition and shape of the object in advance, and the second image is obtained for all the obtained images. By using the method, a method of expecting that the degree of similarity of any of the images becomes high can be considered. However, in this method, a large amount of features with a small difference is generated for the same object, which makes management complicated and increases the time required for searching. There is a problem that it is difficult to aggregate results. For this reason, a method for efficiently managing and storing a large amount of feature amounts extracted for the same object is required.

SUMMARY OF THE INVENTION The present invention has been made in view of the above circumstances, and has as its object the object of the real world which may include an object having a different shape or color depending on a composition to be photographed, or an object having a different shape. Realizes information retrieval using the target image input device, extracts and manages and stores various features of various types of objects, such as object color information and shape information, and calculates similarity of objects. An information storage device that can reduce the calculation time required for and can present information on a photographed object to a user;
An object of the present invention is to provide an information search device, an information storage method, an information search method, and a recording medium on which these methods are recorded.

[0009]

In order to achieve the above object, an information storage apparatus according to the present invention comprises: first image input means for inputting two-dimensional image data of an object and an object name; A first feature value extracting means for extracting a region of the object from each two-dimensional image data input by the image input device, and extracting one or more image feature values from the obtained region image data value; Object information input means for inputting object information relating to the object, an object name obtained by the first image input means, an image feature quantity obtained by the first feature quantity extraction means, and an object obtained by the object information input means Object information management / storage means for generating an object object by associating information for each object, and object information storage means for storing the object object generated by the object information management / storage means And wherein the door.

Here, the object information management / storage means comprises:
Means for managing the object objects stored by the object information storage means and searching whether or not an object object including the object name input by the first image input means is stored in the object information storage means; When the object object including the object name input by the first image input unit is stored in the object information storage unit, the plurality of image feature amounts extracted by the first feature amount extraction unit are Means for associating the object with the object searched by the search means.

The object information management / storage means may include, for the object objects stored in the object information storage means, an image feature amount having a high degree of similarity among image feature amounts associated as the same object object. In such a case, it is preferable to have a means for selecting or collecting image feature amounts to be associated with the object object. Further, the object object stored in the object information storage means associates image feature amounts respectively extracted from images obtained by photographing the object under a plurality of conditions with respect to the same object, and information such as text, images, and voices on the object. It is preferred that it is.

On the other hand, the information storage method according to the present invention captures a two-dimensional image of an object and an object name, extracts an area of the object from each of the captured two-dimensional image data, When one or more image features are extracted and object information relating to the photographed object is input, an object object is associated with the captured object name, the extracted image feature amount, and the input object information for each object. It is characterized in that the object is generated and stored.

Further, the above-described recording medium on which the information storage method according to the present invention is recorded can be distributed as a product, and the scope of the rights of the present invention is to realize the information storage method. The present invention extends to a recording medium on which the above program is recorded.

In order to achieve the above object, an information retrieval apparatus according to the present invention comprises: a second image input means for capturing a two-dimensional image including an object to be searched; and a second image input means. Extracting a region of the object from each of the two-dimensional image data input according to the above, extracting one or more image feature values from the obtained image data values, and a second feature extraction device. Using the extracted image feature amount as a key, an object information stored in each of the object information storage devices described above is searched, and object information search means for searching for desired object information is obtained. And a search result output means for presenting the object information to the user.

Here, the object information retrieving means uses the one or more image feature quantities extracted by the second feature quantity extracting means as a key to retrieve a plurality of object objects contained in the object object stored by the object information storage apparatus. It is preferable to have a means for searching for the image feature amount of (1) and obtaining, as a search result, an object name and object information associated with the image feature amount having a high similarity.

On the other hand, the information retrieval method according to the present invention captures a two-dimensional image including an object to be input for retrieval, extracts an object region from each of the captured two-dimensional image data,
At least one image feature value is extracted from the obtained image data value, and the extracted image feature value is used as a key.
5. The object information storage device according to any one of the items 4, wherein the object information storage device searches for an object object, searches for desired object information, and presents the searched object information to a user.

Further, the recording medium on which the information search method according to the present invention is recorded as described above can be distributed as a product, and the scope of the rights of the present invention is to realize the information search method. The present invention extends to a recording medium on which the above program is recorded.

According to the present invention configured as described above,
An information storage and retrieval method using an image input device for an object having a shape, a color, or the like depending on a composition to be photographed or an object in the real world that may include an object having a change in shape is realized. Extracts various features such as color information, shape information, etc., efficiently manages and stores them, reduces the calculation time required to calculate the similarity of the object, and presents information on the photographed object to the user. There is an effect that can be.

As described above, a recording medium on which the information storage and / or retrieval method according to the present invention is recorded can be distributed as a product, and the scope of the present invention is limited to the information retrieval and / or retrieval method. Alternatively, the present invention extends to a recording medium on which a program for realizing a search method is recorded.

[0020]

Embodiments of the present invention will be described below in detail with reference to preferred embodiments shown in the drawings.

FIG. 1 is a block diagram showing an information retrieval apparatus according to one embodiment of the present invention. In FIG. 1, the information search device according to the present embodiment includes an object information storage unit, an object information storage unit, and an object information search unit. The object information storage unit is a module that associates, as an object object, object information relating to a plurality of objects and feature amounts obtained from the objects, and the object information storage unit is a module that stores the associated object objects of the object information storage unit. Yes, the object information search unit obtains a feature amount from the image based on the key image prepared by the user, and performs a similarity search on the object object stored in the object information storage unit to obtain a similarity to the key image. It is a module that presents information of an object having the obtained feature amount to the user.

The object information storage unit includes a first image input unit 1
01, a first feature amount extraction unit 102, an object information input unit 103, and an object information management / storage unit 104. The object information storage unit includes an object information storage unit 201. The object information search unit includes an image input unit 301, a feature amount extraction unit 302, an object information search unit 303, and a search result output unit 302.

First image input means 10 of object information storage unit
1 acquires an image of an object captured by an image input device such as a digital camera or an image scanner as image data. Further, the first feature amount extracting means 102
Are obtained from the image data acquired by the first image input means 101, a plurality of types of feature amounts regarding the object portion, for example,
Calculate color tone, lightness, saturation, etc. Object information input means 10
3 inputs information on the object. The object information management / storage unit 104 includes, for each object to be photographed, information on the object and a plurality of objects calculated by the first feature amount extraction unit 102 obtained from images of the same type of object photographed under different conditions. An object is generated by associating the type of feature with the feature.

The object information storage means 201 of the object information storage unit
Is the object information management / storage means 1 in the object information storage unit.
04 is stored. The second image input unit 301 of the object information search unit is operated by an image input device such as a digital camera or an image scanner.
An image of the object photographed by the user is acquired as image data. Further, the second feature amount extraction unit 302 calculates a plurality of types of feature amounts regarding the object portion from the image data acquired by the image input unit 301. Object information search means 3
03 acquires object information from the object objects stored in the object information storage unit 201 in the object information storage unit, using the plurality of types of feature amounts calculated by the second feature amount extraction unit 302 as search keys. Search result output means 304
Finally presents the object information obtained by the object information search means 303 to the user.

Hereinafter, a method for generating an object will be described with reference to FIG. FIG. 2 is a flowchart illustrating a flow of a process of the object information storage unit. An object to be searched is photographed in advance by a photographing device such as a digital camera or a video camera. Then, an image captured by the above-described device such as a digital camera or a video camera is input by the first image input unit 101, and an object name is input by the object information input unit 103 (step 1).
1).

From the image data input by the first image input means 101, the first feature amount extracting means 102
After performing smoothing or calibration as necessary, contour extraction is performed, and only the part of the object where the image is captured is cut out. From the image in which the object portion is cut out, the first feature amount extracting means 102 outputs, for example, RG
A feature amount such as the number of appearance pixels or the outline shape of each of B is extracted (step 12).

The object information management / storage means 104 determines whether or not the image being processed is the first object name based on the object name input in step 11 in the object information storage unit. The object object 201 is searched for and examined (step 13). If there is no object object including the input object name and the object name is the first object name, the object information input unit 103 outputs the object information (eg, the amount of money (price), the main cooking method, HTML text such as video, etc.). ) (Step 14), the object information management / storage means 104 generates a new object and associates the object information with the object name (step 1).
5).

If an object object including the input object name exists and is not the first object name,
The object information management / storage unit 104 obtains a similarity between the feature amount associated with the existing object object and the same feature amount extracted in step 12 (step 16). In step 17, it is checked whether or not the similarity is equal to or greater than a predetermined value. If the similarity is less than the predetermined value, the process ends. The object information management / storage unit 104 stores the plurality of feature amounts extracted in step 13 as a subclass of the object object in association with the object information and the object name (step 1).
8).

By repeating the above-described processing for a plurality of objects using images captured under a plurality of conditions, respectively, as input, an object object database accumulated by the object information accumulation means 201 is constructed.

FIGS. 3A and 3B are diagrams showing the structure of the object stored in the above-mentioned object information storage means 201. FIG. 3A is a conceptual diagram (image name: for example, image file name), and FIG. It is expressed like an object-oriented database. In the object information storage unit 201, the object objects are stored in a hierarchical structure as illustrated in FIG. 3, for example.

FIG. 4 is a flowchart showing the flow of processing in the object information search unit. The user of the information retrieval apparatus photographs an object serving as a retrieval key using a photographing device such as a digital camera or a video camera (step 21).
From the image data input by the second image input means 301, after smoothing or calibration is performed as necessary by the second feature amount extracting means 302, contour extraction is performed, and the object is photographed. Only the part that is is cut out. From the image from which the object portion has been cut out, a plurality of feature amounts, such as the number of appearing pixels of RGB or the outline shape, are extracted by the second feature amount extracting means 102 (step 22).

The object information retrieving means 303 retrieves an object name having a similarity of a certain value or more by using a plurality of feature amounts extracted by the feature amount extracting means 302 as a key (step 2).
3) It is checked whether a plurality of the same object names have been determined (step 24). As a result of the search, if the same object name is hit a plurality of times, the search result having a low similarity is deleted (step 25). The object information search means 303 searches object information from the object stored in the object information means 201 based on the object name (step 26). The search result output means 304 outputs, as a search result, an object name and an information name hit in a similar search, for example, to a personal computer (P
The information is presented to the user by displaying it on the display of C) (step 27).

[Embodiment] An object information creator (hereinafter referred to as a creator) firstly asks the user of this search apparatus, for example, vegetables or flowers in a botanical garden in a supermarket, that he wants to know something about it. Is photographed by using a digital camera or the like in a plurality of compositions. Hereinafter, as a specific example, a description will be given of a processing flow when information on a half cut watermelon is stored using the object information storage unit.

The creator photographs the watermelon cut in half from the cut side, and inputs the photographed watermelon to the computer by the first image input means 101 together with the name of the photographed watermelon. The first image input unit 101 assigns a uniquely determined image ID such as a numeric string generated based on the input date, month, day, hour, and second to the input image, for example, a 640 × 480 × 24 bit. The image data is converted into image data handled by the first feature amount extraction unit 102 in a bitmap format or the like.

The first feature amount extracting means 102 receives the image data to which the image ID is assigned from the first image input means 101 and, for example, receives a dynamic contour model (Snakes),
Using a pixel combining method or an object extraction method in an image such as a histogram method, a portion where a watermelon appears in the image is cut out. In this connection, for example, a method described in the following document can be effectively used. (1) "Snakes: Present, Past, and Future" by Mino and Amano (IEICE PRMU97-184 (1997-12), pp81-88) (2) "Image Analysis Handbook" by Takagi and Shimoda (supervised) ( The University of Tokyo Press, 1991.pp.692-693)

The first feature amount extracting means 102 converts, for example, 8 bits for each of RGB colors from the watermelon image.
With respect to the pixel value composed of the following, when the value is 0-127, 0, and when the value is 128-255, 1, the value is (R,
G, B) are (0, 0, 0), (1, 0, 0),
(0,1,0), (0,0,1), (1,1,0),
A cumulative frequency of any of (1,0,1), (0,1,1), and (1,1,1) is obtained, and this is converted into 8-dimensional vector data.

The object information management / storage unit 103 manages an object-oriented database corresponding to the object information storage unit 201 in the object information storage unit, and handles a data structure as shown in FIG. 5, for example. First, the object information management / storage unit 103 checks whether or not the object object including the object name watermelon is already stored in the database.

If the object name watermelon does not exist in the database, the creator is requested to input information about the watermelon. In the object information input means 104, the user describes information about the watermelon such as the name of the main dish using the watermelon such as the price or fruit punch or the image of the watermelon using a format such as an HTML document. , As object information. The object information management / storage unit 103 newly creates an object object of the object class and stores the object name watermelon and the object information in the object.
After storing in the ct object, the image object Image and the image feature amount Image are newly set as child objects.
A feature is created, and the extracted feature amount of the feature amount extracting unit 102 and the image ID are stored in association with each other.

On the other hand, when there is an object object “watermelon” including an object name watermelon, the feature amount extracted by the feature amount extracting means 102 and the image feature amount object are extracted for all image feature amount objects which are child objects of the object object. Feature amount RGBhist stored in
Is calculated, and if the similarity is equal to or greater than a certain value, the process ends.

At this time, the similarity between the feature amount A and the feature amount B is, for example, a generalized ellipse distance (generalized ellipsois).
d distance: T. Seidl and HP Kriegel, "Effci
ent User-Adaptable Similarity Search in Large
Multimedia Databases ”in Proc. Of VLDB,
pp. 506-515, Athens, Greece, Aug. 1997)

(Equation 1) (At this time, s represents the similarity, d represents the generalized elliptical distance), etc., and is usually obtained by 0 <s <1 (1 is most similar, and 0 is most distant. ).

If the similarity is less than a certain value, the image object Image and the image feature amount object I are newly created as child objects of the object "watermelon".
A mageFeature is added, and the plurality of feature amounts extracted by the first feature amount extraction unit 102 and the image ID are stored in association with each other. In addition to the above processing, for example, an image of watermelon taken from the opposite side of the cut, or an image viewed from the side, etc.
An object object related to a watermelon is created by performing the process on an image captured from another angle.

Similarly, for fruits and vegetables other than watermelon, a plurality of object objects corresponding to a multimedia pictorial book are created by repeating the process of storing the characteristic values. Hereinafter, as an example, a flow of processing when a user uses the object information search unit to obtain information about watermelon will be described. In addition, the object information storage unit previously stores, for example, watermelon, apple, and cucumber in whole and in half, respectively.
It is assumed that images captured from various angles are input and stored in the object information storage unit 201 of the object information storage unit.

The user shoots watermelon from an arbitrary direction using an image input device such as a digital camera, and inputs the watermelon to the computer by the image input means 301. The image is converted into image data, such as a 640 × 480 × 24 bit bitmap format, that can be handled by the feature amount extraction unit 302. Similar to the first feature amount extracting unit 102 in the object information storage unit, the feature amount extracting unit 302 receives image data from the image input unit 301 and, for example, receives the above-described active contour model (Snakes), a pixel combination method, a histogram, or the like. The portion of the image where the watermelon is shown is cut out using the in-image object extraction method.

The feature quantity extracting means 302 extracts feature quantities in the same manner as the first feature quantity extracting means 102 in the above-mentioned object information storage section, and extracts eight-dimensional vector data. The object information search means 303 searches all the object objects stored in the object information storage means 201 of the object information storage unit using the feature amount extracted by the feature amount extraction means 302 as a search key, and stores the object information. The similarity is calculated using the same method as that of the object information management / storage means 103 of the part, and the image ID from which the feature amount having a high similarity is extracted and the similarity are obtained. The similarity and the parent class of the image ID As a result.

Here, for example, a threshold value is determined in advance for the similarity, and an object name having a higher similarity is set as an object name having a higher similarity. At this time, the same threshold value is used as the threshold value used when the object information storage unit stores the object object (FIG. 2: Step 17). When a plurality of identical object names are obtained as a result of the search, the highest similarity is regarded as the similarity of the object, and other similarities are excluded. For example, (watermelon, 0.678), (watermelon,
0.328), (watermelon, 0.563), (melon,
0.834), (melon, 0.463), (watermelon, 0.678) and (melon,
0.834) is a valid result.

For example, the image input by the user is a photograph of the surface of a watermelon composed of green ground and black stripes.
It is assumed that the feature quantity extracted from the image has a large value of the vector elements (0, 0, 0) and (0, 1, 0) in the eight-dimensional vector data. In this case, among the image objects of the object objects stored in the object information storage means, (0,
The degree of similarity to a feature amount having a tendency to have large values of the vector elements (0, 0) and (0, 1, 0) increases, and for example, the degree of similarity of an image obtained by capturing the surface of a watermelon or cucumber increases. On the other hand, even if an image of the same watermelon is captured, for example, an image of a cut watermelon cut in half contains a large amount of red, and thus has a low degree of similarity.

Conversely, for example, the image input by the user is a photograph of a red cut of a watermelon cut in half, and the feature amount extracted from the image is (1, 0) of the eight-dimensional vector data. , 0) and (1, 1, 1) have large values of vector elements. In this case, of the image objects of the object objects stored in the object information storage unit, (1, 0,
The degree of similarity to the feature quantity having a tendency to have large values of the vector elements of (0) and (1,1,1) is increased, and for example, an image of the surface of a red apple or a cut of a watermelon cut in half is shot. The similarity of the obtained image is increased.

Therefore, regardless of whether the image input by the user is an image of the surface of a watermelon or an image of a cut, the similarity of any of the image objects included in the object “watermelon” Is higher,
The watermelon that is the desired search result will be returned as one of the search results. The object information search unit 303 extracts object information object information associated with the object name object name from the object objects stored in the object information storage unit 201 based on the valid object name. The search result output means 304 is based on the search results of the object information search means 303 and, in this case, information on watermelon and melon, for example, HT such as amount, main cooking method, video, etc.
The ML document is presented to the user using a method such as displaying the ML document on the surface of the information terminal.

According to the above embodiment, information storage and retrieval using an image input device for an object having a shape, a color, or the like depending on a composition to be photographed, or an object in the real world which may include an object having a change in shape. , It becomes possible to extract, store, manage and store various features such as color information and shape information of an object with respect to various kinds of objects efficiently, and to calculate the time required to calculate the similarity of the objects. And an information storage and information retrieval technique capable of presenting information about a photographed object to a user. Note that the above embodiments are merely examples of the present invention, and the present invention is not limited to these embodiments.

[0050]

As described above in detail, according to the present invention, a plurality of images are created for one object object, and the plurality of images are aggregated to select only those that are significant. Since the feature values related to the object are extracted from these multiple images and stored together with the images based on them, the shape, color, etc. differ depending on the composition to be shot, or the shape changes. Realize information retrieval using image input equipment targeting real world objects that may include objects that
By extracting various features such as color information and shape information of objects and collecting them as appropriate, efficient accumulation / management /
In addition to being able to store the information, it is possible to reduce the calculation time required for calculating the similarity of the object, and to provide the user with information on the photographed object.

[Brief description of the drawings]

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

FIG. 2 is a flowchart showing a flow of processing of an object information storage unit.

3A and 3B are diagrams for explaining the concept of an object configuration, in which FIG. 3A is a conceptual diagram, and FIG. 3B is a diagram representing this in an object-oriented database.

FIG. 4 is a flowchart showing a flow of processing of an object information search unit.

FIG. 5 is a diagram illustrating an example of a data structure of an object stored in an object information storage unit.

DESCRIPTION OF SYMBOLS 101 First image input means 102 First feature value extraction means 103 Object information input means 104 Object information management / storage means 201 Object information storage means 301 Second image input means 302 Second feature value Extraction means 303 Object information search means 304 Search result output means

Claims (10)

[Claims] 1. First image input means for inputting two-dimensional image data of an object and an object name, and extracting an area of the object from each two-dimensional image data input by the first image input means. A first feature amount extraction unit for extracting one or more image feature amounts from the obtained region image data value; an object information input unit for inputting object information on a photographed object; and the first image input unit Object name obtained by
Object information management and storage means for generating an object object by associating the image feature amount obtained by the feature amount extraction means with the object information obtained by the object information input means for each object, and the object information management and storage means An information storage device, comprising: object information storage means for storing a selected object object. 2. The object information management / storage means manages an object object stored by the object information storage means,
Means for searching whether or not an object object including the object name input by the first image input means is stored in the object information storage means, and including an object name input by the first image input means When an object object is stored in the object information storage unit, the plurality of image feature amounts extracted by the first feature amount extraction unit are
2. The information storage device according to claim 1, further comprising: means for associating the object with the object searched by the search means. 3. The object information managing / storing means includes, for an object object stored in the object information storing means, an image feature amount having a high similarity among image feature amounts associated as the same object object. If you do
2. The image processing apparatus according to claim 1, further comprising: means for selecting or aggregating image features associated with the object.
Or the information storage device according to 2. 4. An object object stored in the object information storage means includes: image feature amounts extracted from images of the same object taken under a plurality of conditions;
4. The information storage device according to claim 3, wherein information such as a text, an image, and a sound related to the object is associated. 5. A two-dimensional image obtained by capturing an object and an object name are taken in, an area of the object is extracted from each of the taken two-dimensional image data, and one or more image feature values are extracted from the obtained area image data value. When object information relating to a photographed object is input, an object object is generated by associating the captured object name, the extracted image feature quantity and the input object information for each object, and accumulating the object object. An information storage method characterized by: 6. A recording medium on which a program for realizing the information storage method according to claim 5 is recorded. 7. A second image input means for capturing a two-dimensional image including an object to be input for a search, and extracting an area of the object from each of the two-dimensional image data input by the second image input means. A second feature amount extracting means for extracting one or more image feature amounts from the obtained image data value; and using the image feature amount extracted by the second feature extracting portion as a key. 4. searching for the object object stored by the object information storage device according to any one of 4.
An information search device comprising: an object information search unit that searches for desired object information; and a search result output unit that presents object information obtained by the object information search unit to a user. 8. The method according to claim 8, wherein the object information search unit uses the one or more image feature amounts extracted by the second feature amount extraction unit as a key to store a plurality of image objects included in the object object stored by the object information storage device. 8. The information retrieval apparatus according to claim 7, further comprising means for retrieving an image feature amount and obtaining, as a search result, an object name and object information associated with the image feature amount having a high degree of similarity. 9. A two-dimensional image including an object to be input for retrieval is fetched, an area of the object is extracted from each fetched two-dimensional image data, and one or more image feature values are extracted from the obtained image data value. And using the extracted image feature amount as a key to search for an object stored by the object information storage device according to any one of claims 1 to 4, and search for desired object information. An information retrieval method, characterized by presenting selected object information to a user. 10. A recording medium on which a program for realizing the information search method according to claim 9 is recorded.