JPH09265472A - Picture database system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a picture database system capable of retrieving a picture by a character string keyword even when the character string keyword is not registered in the picture and automatically registering a keyword in a picture file. SOLUTION: A picture database 3 stores a picture file to be retrieved and a keyword picture file 4 is a dictionary file and stores each picture and its corresponding character string keyword as a pair. When a character string keyword is inputted from an input means 5, a keyword picture substituting part 6 substitutes a keyword coincident with the character string keyword inputted from the input means 5 by a corresponding picture by the use of the file 4. A picture recognition part 7 recognizes the substituted picture by matching with a picture file stored in the database 3 and retrieves and outputs the picture. The recognition part 7 recognizes also the picture in the file 4 which coincides with the picture stored in the picture file in the database 3 and registers a character string keyword corresponding to the coincident picture in the picture file of the database 3.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image database system for creating a database of image data and searching it.

[0002]

2. Description of the Related Art There is an increasing number of image database systems capable of storing and retrieving image files in comparison with conventional character-based databases. In a character-based database, a keyword for searching is added to each file in the database, a keyword character string of a file to be searched for is input at the time of searching, and a file is searched by matching this character string with the keyword of the database. Also,
Even if the search keyword is not added, since the file itself is composed of the character string, it is possible to search using the character string in the file itself as the keyword.
However, since the image file itself is an image file in the database, it is necessary to determine the character string corresponding to the image and add the character string to the image file when registering the file in the database. Then, the file search of the image database was performed based on the character string. Japanese Unexamined Patent Publication No. 6-282582 discloses a method of automatically identifying and extracting a phrase included in an image and registering it as a character string in a database together with image data. The above-mentioned publication discloses a technique of extracting and recognizing a phrase included in an image and automatically registering it in image data as registration information for the image.

[0003]

The above-mentioned conventional technique has a problem that an image cannot be searched unless a character string is registered as a keyword. In addition, in the registration of the keyword for the image, when the image includes a character, it can be registered as a keyword in the image data by reading the character. However, if there are no characters in the image, there is a problem that nothing can be registered as a keyword.

An object of the present invention is to provide an image database system which enables an image search by a character string keyword even if the character string keyword is not registered in the image.

Another object of the present invention is to provide an image database system for automatically registering keywords in an image file.

[0006]

In order to achieve the above first object, a plurality of image files holding images to be searched are stored in the first storage means, and the image files are stored in the second storage means. A pair of character string keywords corresponding to this image is stored, and an image having a character string keyword that matches the character string keyword input from the input unit by the keyword image replacement unit is output from the second storage unit, The image file holding the image that matches the image output from the keyword image replacement unit by the recognition unit is taken out from the first storage unit.

In order to achieve the above-mentioned second object, a plurality of image files holding images to be searched are stored in the first storage means, and the second storage means corresponds to the images and the images. The character string keyword is stored as a pair, and the recognition unit recognizes the image of the second storage unit that matches the image held in the image file of the first storage unit, and the character string corresponding to the matched image. The keyword is registered in the image file of the first storage means.

[0008]

DESCRIPTION OF THE PREFERRED EMBODIMENTS One embodiment of the present invention will be described below with reference to the drawings.

FIG. 1 shows the configuration of an image retrieval system which is an embodiment of the present invention. The image search system 1 has a control unit 2 that controls the whole, and each unit is connected to the control unit 2 via a bus. The image database 3 stores a plurality of image files holding images to be searched. An example of the image file is shown in FIG. 3 and will be described later. The keyword image file 4 is a so-called dictionary file, and stores partial images that frequently appear in the image database 3 and character string keywords attached to the partial images as a pair. The input means 5 is for inputting a character string keyword for searching. The keyword image replacement unit 6 searches the keyword image file 4 using the character string keyword input from the input means 5,
By outputting an image having a keyword that matches the input character string keyword, the character string keyword and the image are replaced. The image recognition unit 7 uses the image replaced by the keyword image replacement unit 6 to generate the image database 3
Is output, and the image file holding the matching image is output to the output means 8. As a result, the image file holding the image corresponding to the character string keyword input from the input means 5 can be searched and output.

FIG. 2 shows a storage state of each image in the keyword image file 4, which is a character string keyword portion 4
a and the image portion 4b corresponding thereto are stored as a pair. For example, the character string keyword “triangle” is stored as a pair for the image “Δ”, and the character string keyword “mountain” is stored as a pair for the image in the shape of a mountain. . Further, the image and the character string keyword do not necessarily have a one-to-one correspondence, and a plurality of character string keywords “circle, circle” may be stored as a pair for the image “◯”. When the keyword image replacement unit 6 replaces the character string keyword and the image corresponding thereto, the keyword image replacement unit 6 refers to the keyword image file 4 to perform the replacement.

FIG. 3 shows each image file in the image database 3. Each image file consists of an image supplementary information section 3a and an image section 3b. The image supplementary information section 3a is provided with recording fields such as the file name of each image file, the data capacity, the image title, the file format, the search character string keyword, and the registration date, and the image supplementary information section 3b. An image corresponding to the part 3a is stored.

Next, the operation of the present invention will be described with reference to FIG. FIG. 4 shows a flow of processing for searching an image with the character string keyword when the character string keyword is not registered in the image. When searching for an image, first, the input means 5
Input a character string keyword corresponding to the image to be searched with (21). Next, the keyword image replacement unit 6 receives the input character string keyword and the keyword image file 4
Matching process with the character string keyword part 4a of
It is judged whether or not this character string keyword exists in the keyword image file 4 (22). When there is a matching character string keyword, the keyword image replacing unit 6 loads the image corresponding to the matching character string keyword, replaces the input character string keyword with the corresponding image, and outputs (23). . If there is no corresponding character string keyword, the user is asked whether to input another character string keyword (29). If it is input, the process returns to step 21, otherwise, the process ends. Next, the control unit 2 loads an image file from the image database 3 (24), and the image recognition unit 7 loads the keyword image replacement unit 6
Matching recognition of the image output from the image and the image held in this image file is performed (25). In this image matching recognition, an arbitrary partial image of the image file loaded from the image database 3 is cut out, normalized to the same size as the keyword image, and then subjected to image processing such as exclusive OR for each pixel with the keyword image. , Using a method such as calculating the area of the common part. Then, it is determined whether or not the image output from the keyword image replacement unit 6 matches the image file loaded from the image database 3 to be searched (26), and if found, the image file is searched. As a result, it outputs to the output means 8 (27). This keyword image replacement unit 6
A series of search processing for a certain image output from is performed for all the images within the search range in the image database 3 (28). As a result, the image file holding the image corresponding to the character string keyword input from the input means 5 can be searched and output.

FIG. 5 is an example of searching and outputting a person image in an image. First, the character string keyword “person” corresponding to the image to be searched is input by the input means 5 (FIG. 4 (2
1)) and whether the character string keyword “person” exists in the keyword image file 4 (FIG. 2) (FIG. 4 (22)), and if there is, the keyword image replacement unit 6
Outputs a person image corresponding to the character string keyword "person" ((23) in FIG. 4). Next, the image recognition unit 7 performs image matching recognition between the person image 4b and the image 3b of an image file in the image database 3 (FIG. 4).
(25)). As a result of the image matching recognition, when the person image 3c is detected at the lower right of the image 3b, the image file holding the image 3b is output to the output means 8 (FIG. 4).
(27)).

From the viewpoint of the retrieval processing speed, the retrieval using the character string keyword which is generally used is much faster than the retrieval using the above image. for that reason,
If the character string keyword corresponding to the image held in the image file can be easily registered, the search process using the database system can be speeded up.

FIG. 6 shows a flow of processing for registering a character string keyword corresponding to an image in an image file in the image database by using this search system. In FIG. 6, the control unit 2 first loads an image file from the image database 3 (31), and also loads an image from the keyword image file 4 (3).
2). Then, as in step 25 of FIG. 4, the image recognition unit 7 includes the image held in the image file loaded from the image database 3 in (31) and the image matching the image loaded in (32). Image matching recognition of whether or not there is is performed (33). The recognition result is judged (3
4) If the target image is found in the image held by the image file loaded from the image database 3 in (31), the keyword image replacement unit replaces the image loaded in (32) with the character string keyword ( 35). And
The control unit 2 sends the character string keyword to the image database, and writes and registers the character string keyword in the search character string keyword column of the image accompanying information unit of the recognition target image file (36).
If the target image is not found in the image held in the image file in (34), it is judged whether or not there is an image that has not been matched in the images in the keyword image file 4 (37). If so, the process returns to (32) to load another image, and if not, the process ends.

The keyword generation / registration process shown in FIG.
It may be performed at the time of image registration, or may be performed collectively for all the image files when the image files are registered without keywords and a certain amount of the image files are registered in the database. Further, a plurality of keywords may be registered for one image file so that the image search can be easily performed. In the example of FIG. 2, the shape of the image is used as the character string keyword, but the invention is not limited to this. For example, instead of the shape of the image, the color and other characteristics may be associated with the image as a character string keyword. Further, the image registered in the keyword image file 4 may not be the image itself, but the definition for recognizing the image may be registered. For example, if a character string keyword such as “black” may be input to search for a black image instead of the shape of the image as the character string keyword, the image part corresponding to the character string keyword “black” is input. To "90% of the black part in a certain area
The definition of “determining the above as black” is stored. When the character string keyword “black” is input from the input means 5, the image recognition unit 7 performs image processing for checking the ratio of the black portion to the image to be searched according to the definition stored in the keyword image file 4. When the black portion is 90% or more, it is determined that the image includes “black”, and the image is output. Furthermore, the image registered in the keyword image file 4 is not the image itself,
A recognition program executed by the image recognition unit 7 may be used. For example, it is a recognition program for recognizing a red color portion, taking an outline of an image, and the like. String keyword "red"
Assuming that a “recognition program for recognizing a red color portion” is stored in the image portion corresponding to, when the character string keyword “red” is input from the input means 5, the image recognition unit 7 causes the character string keyword. The recognition processing is executed according to the recognition program stored in the image portion corresponding to "red", and the image file including the red portion is output. In any case, correspondence with the character string keyword is necessary.

FIG. 7 shows a state in which the character string keyword corresponding to the image is registered in the image file in the image database 3 by performing the processing shown in FIG. When registering a character string keyword, an image file is loaded from the image database 3, and then the keyword image file 4 is loaded.
A certain image is loaded, and image matching recognition between that image and the image held in the image file is performed (Fig. 6 (3
3)). If the images match as a result of the image matching recognition, the keyword corresponding to the images is registered in the image file loaded from the image database 3 (FIG. 6 (36)). Thereby, the character string keyword can be automatically registered in the image file.

The above is an embodiment relating to an image database of still images, but regarding moving images, each frame of the moving image is regarded as a still image, and each frame is stored in the first storage means. , The above image search system can be used for moving images as it is. In that case, since the number of frames increases when the moving image is divided into each frame, a method of storing a part of the frames may be considered.

The input means for inputting the keyword is
Generally, many keyboards are used, but the image search system may be connected to a network connection and input from a client terminal or another network system via a line.

[0020]

In an image database system capable of filing and searching a large number of images, the user can search for a target image having the image feature by inputting the image feature as a character string keyword. Further, in registering the image file in the database, the image recognition unit recognizes the image feature and registers the keyword representing the image feature in the image file without the need for the registrant to consider and register an appropriate keyword in the image file. Therefore, the database registration work of the registrant can be reduced.

[Brief description of drawings]

FIG. 1 is a diagram showing a system configuration according to an embodiment of the present invention.

FIG. 2 is a diagram showing a specific example of a keyword image file in FIG.

FIG. 3 is a diagram showing a specific example of the image database of FIG.

FIG. 4 is a diagram showing a flow of image search processing using a character string keyword in the system of FIG.

FIG. 5 is a diagram illustrating a specific example of image search processing.

6 is a diagram showing a flow of a character string keyword registration flow process in the system of FIG.

FIG. 7 is a diagram illustrating an example of a character string keyword registration process.

[Explanation of symbols]

1: Image retrieval system, 3: Image database, 4: Image file for keyword, 5: Input means, 6: Keyword image replacement section, 7: Image recognition section, 8: Output means

Claims (2)

[Claims] 1. A first storage means for storing a plurality of image files holding images to be searched and a second storage means for pairing an image and a character string keyword corresponding to the image.
Storage means, input means for inputting a character string keyword, and keyword image replacing means for outputting an image having a character string keyword matching the character string keyword input by the input means from the second storage means. An image file holding an image matching the image output from the keyword image replacing means, and recognizing means for taking out the image file from the first storage means; and an image file having an image corresponding to the input character string keyword An image database system characterized by searching from a second storage means and outputting. 2. A first storage means for storing a plurality of image files holding images to be searched, a second storage means for storing the image and a keyword corresponding to the image as a pair, and A recognition means for recognizing an image of the second storage means that matches the image held in the image file of the first storage means, and registering a keyword corresponding to the matched image in the image file of the first storage means; An image database system comprising: