JPH02253373A - Preparation method for picture data base - Google Patents

Abstract

PURPOSE:To additionally register or correct a key word by instructing a high frequency input key word and an important key word to the picture data inputted in the past by means of the preceding retrieving history. CONSTITUTION:A retrieving history information control part 8 which controls the retrieving key word retrieving formulas inputted at execution of preceding retrieving actions as well as the input frequency, the success of fitness, and the latest input time of those retrieving formulas, a retrieving history information storing part 9 which stores the retrieving key word retrieving formulas and their input frequency, success of fitness, and latest input time are provided. Then a high frequency input key word and an important key word are shown via the part 9 to the picture data inputted in the past. As a result, the key words can be additionally registered and corrected and the fitness rate is extremely improved for a picture data base at retrieving.

Description

[Detailed Description of the Invention] [Industrial Application Field] The present invention relates to an image database creation method for effectively assigning and managing keywords in an image database search system using a search key management/verification method. .

[Conventional technology]

Conventionally, as a search method for image databases, similar to text database search methods, only keywords representing the content of images are managed, and the search key management/verification method is used to search for desired image data by inputting a keyword search formula. Another method is to search for and narrow down desired image data by repeating the display of several representative plane images and instructions from the user.

The former method is suitable for target images that can clearly express and describe the content recalled by the user. The latter method is suitable for target images in which it is difficult to clearly express the user's image.

FIG. 6 shows a conventional example of an image database search system using the former search key management/verification method.

In the figure, 1 is a scanner section, 2 is a keyword input section, 3 is an image data storage section, 4 is a keyword storage section, 5 is an image database section, 6 is a search operation section, and 7 is a search collation section.

An overview of the operation of this system is shown below.

When registering image data of a vase as shown in FIG. 7, first, an original image is captured by the scanner unit 1, compressed using various encoding methods, and stored in the image data storage unit 3. Next, a word (keyword) expressing the feature of the contents regarding the vase is inputted from the keyword input section 2, and is registered in the keyword storage section 4, and the correspondence with the stored image data is specified. For example, the keywords are “round shape” and “Jomon”.
``Register red soil j, etc.'' There is also a method of managing keywords in the keyword storage section 4 to improve the search speed by creating a keyword reverse file.

Next, in order to search for image data from the image database unit 5 registered in the above manner, the operator inputs an appropriate keyword search formula that has been recalled corresponding to the image data to be searched from the search operation unit 6. The search matching unit 7 compares the input keyword with the keyword information in the keyword storage unit 4, searches for image data (there may be more than one) that matches the input keyword search formula, and performs the search operation. Displayed in section 6. In addition, in order to perform a wide search based on [content] including synonyms, broader terms, narrower terms, synonyms, etc. of the entered keyword, there is also a method of expanding word weight using a thesaurus.

[Problem to be solved by the invention]

In the case of the conventional image database search method shown above,
At best, corresponding image data can be extracted only when a search is performed using a pre-registered keyword or, at most, its synonyms, higher rank, lower rank, or similar synonyms.

When searching with a concept other than the keyword [round shape] [Jomon ko [red clay]] and its synonyms, superordinates, subordinates, and synonyms (for example, [handle]) registered in the keyword storage unit 4 for the vase shown in FIG. ), the drawback was that even though corresponding image data existed, it was not possible to search for it.

In order to solve the above-mentioned drawbacks, this invention makes it possible to add or modify keywords by presenting frequently input keywords and important keywords using previous search history for image data input in the past. The present invention provides a method for creating an image database using the following methods.

(Means for Solving the Problems) An image database creation method according to the present invention provides information on each search keyword search formula inputted during a previous search execution, the number of inputs thereof, success or failure of matching, and information on the conventional image database system. Similar to the search history information management section that manages the latest input time, the search keyword search expressions and the number of times they were input.

A search history information storage unit is provided to store the success/failure of matching and the latest input time, and keywords can be identified by referring to the search history information storage unit and presenting frequently input keywords and important keywords for image data input in the past. It is designed to allow additional registration or modification.

[Effect]

In this invention, since frequently input keywords and important keywords are presented from the search history information storage section for image data input in the past, keywords can be additionally registered or modified based on these.

〔Example〕

FIG. 1 is a block diagram of an apparatus for explaining an image database creation method according to an embodiment of the present invention. In the figure, 1 to 7
is the same as in Figure 6. 8 is a search history information management department;
9 is a search history information storage section.

Next, the operation will be explained. First, a procedure for searching for desired image data from the image database section 5 using a keyword designation method will be explained according to the processing flow shown in FIG.

Note that (1) to (6) indicate each step.

Enter the appropriate keyword search formula you have recalled from the search operation section 6 (1). The search and collation unit 7 collates the input keyword with the keyword file in the keyword storage unit 4 (2), and makes a determination (3). If the matching is successful, the corresponding image data is extracted; if the matching is unsuccessful, the information "No matching image data" is displayed on the search operation unit 6 (4).

(5). Finally, the history of search results is recorded and updated on the search history information storage section 9.

FIG. 3 shows the contents of the search history information storage section 9.

For each search keyword previously entered, the search expression and number of times it has been entered so far are shown here.

The presence or absence of relevant data and the latest search time are entered and updated.

In FIG. 3, symbols A, B, and C are respectively input keywords (for example, [pot], [Jomon], [handle, etc.), and "*" is a logical product. [A*B*C] and [A*
C] had no matching data despite the large number of inputs. [A*B] had a large number of inputs and had matching data. In other words, [A] and [B] are keywords that are important in the stored image data and have already been assigned, and [C] is assigned to the stored image data. Although the keyword is not listed, it is considered to be an important keyword and should be added as a keyword. On the other hand, since [B*C] is input less frequently, this indicates that there is no great interest in the search results obtained by the combination of [B] and [C].

Next, a procedure for adding and correcting keywords to image data input in the past using search history information will be explained according to the processing flow shown in FIG. Note that (1) to (7) indicate each step.

The search history information management section 8 compares the search history information with the search history information registered in the search history information storage section 9 and (1) displays a message promoting appropriate keyword registration on the keyword input section 2 (2). The operator can know additional required keywords through promotional messages. Specify a subset of the image database to which the keyword is assigned
(3). Next, image data candidates included in the specified subset are sequentially displayed (4). The operator uses a mouse or the like to re-enter the image data candidates to which the keyword should be added (5), and then inputs the keyword into the keyword storage section 4.
(6). Repeat the above steps until there is no more image data.

The details of the promotion message will be explained according to the display example shown in FIG. In the figure, the underlined portions indicate information input by the operator from the keyword input section, and the others indicate messages from the system side.

First, the system side displays the contents of the search history information storage section 9. Here, the keyword C [handle] has not been added to the previously accumulated image data, and the keyword A [handle] has not been added to the previously stored image data.
[Tsubo] and B [Jomon] have already been assigned, and the operator is notified that they are important keywords. The operator knows from this message that it is necessary to add the keyword C [handle] to the corresponding image data. Next, the search history information management unit 8 uses the keyword C[
A candidate set of image data that needs to be attached with [handle] is presented. Here, the candidate set is a set of search expressions that are input with keywords that have been input more frequently than in the past search history and that have been pointed out by the system as needing to be added ([C] in Figure 5). It is shown as the sum of the search expressions excluding the keyword part. In the example in Figure 5, subtract [C] from [A*B*C] and [A*C] <[A+A*B]
; Indicated as [A]. However, [A] or [A*
The operator determines a subset of the image data according to the number of image data to which the keyword B] is assigned. On the system side, the operator who sequentially displays the image data of the corresponding subset can additionally register a new keyword C by instructing the image data to which the keyword C should be added (for example, by clicking the mouse). can.

It goes without saying that the present invention can be applied not only to image databases but also to text databases in a similar manner.

(Effects of the Invention) As explained above, the present invention provides an image database search system based on keyword input, which includes a search history information management section and a search information storage section, and performs appropriate processing for previously inputted image databases according to the previous search history. By facilitating the input of keywords, the system has the advantage that missing keywords and important keywords can be additionally registered, thereby significantly improving the precision when searching an image database. Furthermore, it has the advantage of being able to dynamically follow changes in the interest of the search target over time.

[Brief explanation of drawings]

FIG. 1 is a system configuration diagram of an embodiment of this invention, FIG.
FIG. 4 is a processing flow diagram of this embodiment, FIG. 3 is a diagram showing the contents of the search/away information storage section, FIG. 5 is a diagram showing an example of an input prompt message, and FIG. 6 is a diagram of the conventional embodiment. The system configuration diagram, Figure 347, is a diagram showing an example of input image data "pot". In the figure, 1 is a scanner section, 2 is a keyword input section, 3 is an image data storage section, 4 is a keyword storage section, 5 is an image database section, 6 is a search operation section, 7 is a search matching section, and 8 is search history information The management section 9 is a search history information storage section. Figure 1 Figure 2 Figure 3 Figure 6: Search operation section Figure Figure 6

Claims (1)

[Claims] a scanner section for inputting image data; an image data storage section for storing the image data; a keyword input section for inputting search keywords for the image data; a keyword storage section for storing and managing the search keyword information; a search operation section that displays search results for search keywords;
In an image database system comprising a search matching unit that matches the search keyword with the keyword storage unit, each search keyword search formula input during a previous search execution, the number of inputs thereof, success or failure of matching,
Similar to the search history information management section that manages the latest input time, there is also a search history information storage section that stores the above-mentioned search keyword search expressions, the number of times they have been input, success/failure of matching, and the latest input time. On the other hand, an image database creation method characterized in that keywords are additionally registered or modified by referring to the search history information storage section and presenting frequently input keywords and important keywords.