JPH02220177A - Picture retrieving system - Google Patents

Abstract

PURPOSE:To facilitate the change of a retriever by comparing a stored and held index with a produced retriever and changing the index added to a stored picture to prepare a new retriever. CONSTITUTION:When 'being square' is inputted via an interface (a) to the part of an ear of a coffee cup, i.e., a three-dimensional index which is newly received by a retriever preparing device 2, the device 2 sends the words 'being square' to a word-function description collating device 6 via an interface (g). The device 6 collates a function expression corresponding to the words 'being square' at a data set 7 via an interface (h) and sends a command to the device 2 to reduce a form parameter. The device 2 changes the parameter of a super- secondary function to change a round ear of the coffee cup into the square one. As a result, a retriever is prepared and the attribute of the retriever is change with use of words.

Description

[Detailed Description of the Invention] [Industrial Application Field] The present invention relates to an image search method for an image database, and involves changing an index already attached to a stored image to an attribute similar to a search target object. This relates to an image search method that creates a new search element.

[Traditional technique]

The conventional search method for image databases is to set the attributes of the search object to a "cylindrical red tip" or "radius 10".
You can search by inputting words such as ``Centimeter Kotop'', by inputting a drawing or two-dimensional image, and by understanding the image, or by creating a three-dimensional search element for the search object. There is a way to do it.

[Problem to be solved by the invention]

However, the attributes of the search object (shape, color, texture, etc.)
It is possible to describe accurately in words, to draw the attributes of an object on a drawing for which the specific attributes are unknown, and to use a three-dimensional search engine to specifically describe the detailed attributes of the entire object. was difficult to create.

Another problem is that even if the attributes of the search target are known, creating detailed attributes as search elements requires skill in operations such as conventional mouse changes, which requires training. was there.

[Means to solve the problem]

In order to solve these problems, the present invention aims to facilitate the creation of search elements and search for images by using already accumulated indexes. means for accumulating and retaining indexes of image structures, means for creating search elements,
A means is provided to compare the input search element with the stored index, and the necessary part is selected when searching for the desired image from the index already assigned to the stored image. The most important feature is that it allows the changed part to be changed to a desired attribute.

[Effect]

Thereby, a new search element is created by selecting an attribute or a portion thereof required as a search element from the indexes already assigned to stored images, and making changes to the selected area. Also, when creating a new search child or making changes to the index,
By describing attributes using functional expressions, it becomes possible to create search elements and change attributes using words.

〔Example〕

DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described below with reference to the drawings using three-dimensional image data as an example. In the embodiment, the similarity of shapes is compared based on the difference in distance between each function parameter of the hyperquadratic function. FIG. 1 is a block diagram showing the configuration of an apparatus according to an embodiment of the present invention, in which 1 is an image database;
2 is a search element creation device; 3 is a search element-index collation device that compares the search element created in device 2 with the index given to the image stored in device 1; 4 and 5 are displays; 6 7 indicates a device that matches words expressing attributes of objects with expressions using functions, 7 indicates a data set expressing attributes of objects using functions, a indicates an interface between the user and the device, and b-'-g indicates the device. shows the interface between

FIG. 2 is an explanatory diagram when the hyperquadratic function used in the specific example is expressed in spherical coordinates, showing the magnitude and direction of the vector X, and 10 is the surface of the hyperquadratic function. Moreover, the hyperquadratic function in this case is shown by the following formula.

Here, η and ω indicate latitude and longitude, respectively, and have the following ranges.

O≦ω×2π Further, the coefficients alSa2 and a3 are scale parameters in each axis direction and are scalars. ε1 and ε2 are shape parameters, which are powers of real numbers. Then, the shape can be changed in two axes by ε1, and the shape can be changed in the xy plane by ε2.

FIG. 3 shows an example of a search element that is first input.

Further, FIG. 4 shows an index assigned to an image as a search result, and FIG. 5 shows a new search element created by changing the index shown in FIG. 4.

Referring now to FIG. 1, an example of the operation will be described below, which is an embodiment of the present invention, in which a coffee cup is considered as a search object. The user provides the word "angular" to the search element creation device 2 via the interface a.The search element creation device 2 inputs the expression "angular, kotsupu" via the interface g to the word-function matching device. Transfer to 6.

The word-function matching device 6 refers to the data set 7 via the interface h. In dataset 7, 4
(“angular” 11 Shape parameter ε2≦0.3 (default value ε2・0.
)) (“Koppu” 11 Size (def*I$) value); al=a2=a3=
I.Q.

Shape parameter εl≦0.5.62 is arbitrary (default value: 61=0.1, ε2-1.0)
). Here, the range of shape parameters in the description of the tip is limited because the tip needs to have a stable shape when placed, and this constraint condition is expressed by parameters. words-
In the function matching device 6, the function parameter ε1 is determined by adopting default values for parts that are not described in words.
・0.1. ε2・0.1. al・a2=a3・10.0
is sent to the search element creation device 2, and a general shape of the search object as shown in FIG. 3 is created using a hyperquadratic function.

The created search element is displayed on the display 5 via the interface f, and is also sent to the search element-index matching device 3 via the interface C. In the search element-index matching bag W3, the input search element, that is, the function parameter of the hyperquadratic function is compared with the ° parameter of the three-dimensional index assigned to the image stored and held in the image database 1. . Then, an image to which a three-dimensional index having a parameter closest to each function parameter of the search child created by the search child creation device 2 is given is displayed on the display 4 via the interface e, and is further displayed in FIG. The three-dimensional index given to the image of the search result as shown in FIG.

As shown in Figure 4, the first input search element (
A three-dimensional index similar to that shown in FIG. 3) with a coffee cup handle added thereto is displayed.

When the user recognizes that the search result image displayed on the display 4 is not the image he/she was searching for, the search element creation device 2 changes each part of the three-dimensional index that was given to the search result image. Among them, the attribute part necessary for searching for a desired image is selected via interface a.For each part of the selected index, the user gives a verbal instruction to change the shape to the one to be searched.0 For example, if you enter "square" via interface a for the handle of a coffee cup, which is a new three-dimensional index sent to the search element creation device 2, the search element creation device 2 inputs the word "square" from the interface g. The word "square" is transferred to the word-function description matching device 6 via the interface g.

The word-function matching device 6 matches the functional expression corresponding to the word "square" in the data set 7 via the interface h, and sends a command to the search element creation device 2 to reduce the shape parameter ε1 by 0.5. send. The search element creation device 2 changes the round-shaped handle into a square-shaped handle as shown in FIG. 5 by changing the parameters of the hyperquadratic function.

Similarly, by inputting the words "slightly rounded cup", the portion of the cup shown in FIG. 3 20 is transformed into a rounded cup as shown in FIG. 4 21. This operation creates a new search object called a coffee cup with a square handle. When the transformation is completed, the image search is performed again using this new search element.

The above operation is performed once, and as the changes to the 3D index are made in detail until the desired image is searched, the search element is gradually described in detail, and the search target image is Image searches will be performed while narrowing down the search results.

In this embodiment, the three-dimensional shape is described using a hyperquadratic function, but any method may be used as long as it can describe the three-dimensional shape of the image structure.

Furthermore, although three-dimensional image data has been described as an example, even two-dimensional image data can be realized more easily by using exactly the same method.

In addition, in this embodiment, the shape of the object is used as an example of an attribute of the object, but there are many other ways to express the attributes of the object, such as colors and textures. A similar effect can be obtained.

〔Effect of the invention〕

As explained above, when creating a search child, the database can provide an index that is a sample of the shape that the user is trying to create, and all you have to do is extract the necessary parts from that index and change each part. You can easily create a search child. When making this change, the user does not need to be skilled in creating search elements because the change is made using verbal instructions.
Search elements can be easily changed.

In addition, the present invention first creates a search element by inputting words for parts that can be easily expressed in words, and then changes the index already assigned to the image and uses the part that is difficult to express in response to a search request. It is possible to gradually describe the search element in detail, and it becomes possible to narrow down the images to be searched.

In addition, in the present invention, when changing the index, by using only the changed part as a search element, it is also possible to perform a search that searches all images that include the attributes of the changed part, that is, a search based on partial attributes. It is.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing the configuration of an embodiment of the present invention, FIG. 2 is a drawing explaining the coordinate system of a hyperquadratic function, and FIG. 3 is a drawing showing an example of a search element that is first input. FIG. 4 is a drawing showing an example of an index given to an image of a search result, and FIG. 5 is a drawing showing an example of a new search element created by changing the index shown in FIG. 4. 1... Image database 2... Search element creation device 3... Search element-index matching device 4.5... Display 6... Word-function description matching device 7... Attribute expression of objects Data set 10 described by functions...Hyperquadratic function coordinate system 20...Example 21 of search child created by device 2...
・Example 22 of a three-dimensional index given to a search result image... Examples of search elements a to h obtained by transforming the three-dimensional index... Interface Figure 1

Claims (1)

[Claims] (1) In an image database image search method, a first means for accumulating and retaining an index of image structures;
The method comprises a second means for creating a search element, and a third means for comparing the index accumulated and held in the first means with the search element created by the second means. An image search method characterized by creating a new search element by changing an index already assigned to an image.