JPH0251775A - Retrieving device for picture information - Google Patents

Abstract

PURPOSE:To enable a user himself/herself to easily set a deciding standard for resemblance for each retrieval by comparing and evaluation the attribute value of the component elements of each picture with the attribute value of the edited component elements based on the estimated value of the resemblance deciding standard. CONSTITUTION:A picture component element calling part 22 stores previously the component elements and their attribute information as the graphic information, etc., representing each component material of the picture information and calls out a desired component element in response to a user's request. A picture component element editing part 23 edits the attribute information on the position, the size and the form of each called picture component element. In this case, a resemblance deciding standard value estimating part 30 estimates the resemblance deciding standard value from the user's editing information obtained at the part 23. Then a candidate picture information evaluating part 25 compares for evaluation the attribute value of the component elements of each picture of a retrieving correspondence list storing part 26 with the attribute value of the edited component elements based on the estimated resemblance deciding standard estimated value. Then the picture information having high resemblance between both attribute values is selected as a retrieving candidate.

Description

Detailed Description of the Invention (Technical field to which 5I! Akira belongs) The present invention is an image information retrieval method that quickly searches and presents desired image information using a simplified image as a key from an image database that stores information on a plurality of images. It is related to the device.

(Prior Art) Conventional techniques for searching desired image information from a single image database include the following.

(a) A uniquely identifiable identification symbol such as a screen number is given to each piece of image information, and as shown in FIG. Accumulation position I in MW medium 1 such as apple 7 and person 8! ! A correspondence list 4 between 2 and identification symbol 3 is prepared in advance.

When the user inputs 5 the identification symbol 3 attached to desired image information, the identification symbol 3 is referred to in the correspondence list 4, and M
A method of presenting M-stacked image information at a corresponding storage position of the storage medium 1, in this case, point b.

(a) A plurality of identification symbol groups are given to each image information in advance, and each image information storage medium 1
Correspondence list 4 between storage position [2 and identification symbol group 3' in
Prepare in advance.

When the user inputs a plurality of identification symbols that are expected to be attached to desired image information, a group of these identification symbols 3'
is referred to on the correspondence list 4, and a plurality of pieces of image information are presented in descending order of the degree of suitability of symbol groups by ra.

In method (a) above, the user must accurately remember the identification symbol attached to each piece of image information, so when the number of pieces of image information stored in one image database becomes enormous, Searching becomes significantly more difficult.

Furthermore, in the method (a) above, words called keywords that represent the content of image information are generally used as identification symbols. By applying this keyword, even if the information necessary to search for desired image information is vague to some extent, it is possible to search for a plurality of pieces of image information that are candidates for desired image information.

However, if only a group of symbols such as keywords are used as search information, there is a high possibility that image information with a layout far different from what the user expected will be presented. In other words, each component (
The constituent elements referred to here are the semantic parts that constitute one image information. For example, for the image information in Figure 10, "
Since the positional relationship, size, posture, etc. of the constituent elements of "desk", "apple", and "person" are information that cannot be specified by keywords, the layout of the image information will be completely different.

In particular, when the number of image information stored in an image database becomes enormous, the number of candidate image information also becomes enormous, making it difficult to perform a quick search. Therefore, a method for searching desired image information from an image database is a method in which the user draws a schematic diagram representing the desired image information, inputs the schematic diagram as search information, and presents image information most similar to the schematic diagram. is desirable.

This search method is called an image similarity search method, and in order to implement it in accordance with JIS-9, it is necessary to automatically recognize and understand the schematic shapes drawn by the user.

This is a technology that is much more difficult to implement than before.

In addition, it is a very time-consuming task for the user to create a schematic shape by hand.

(1) Therefore, instead of identification symbols or keywords, the search information is a key that uses various attributes such as the constituent elements that make up one image information and the size, orientation, posture and positional relationship of each constituent element within the image information. Image search methods have been proposed in the past.

For example, in the image information in Figure 12, the constituent elements are "desk" 6,
"Apple" 7. ``Chair Jll, r window J 12, each of the two components has a size, orientation, posture and two positions.

and attribute information such as mutual positional relationships between components.

The key image search method uses a drawing processor in the form of verbal input when the user inputs the search information. What is the language input format drawing processor? Figure 13 (,)
As shown in , when you first input the name of a component, for example "apple" 7, a simple schematic diagram 1 representing the shape of the component is displayed.
8 is processed by the drawing processor 13, displayed, and the default attribute information 15 regarding the constituent element name 14 is written in the search graphic editing list 14A shown in FIG. That is, the size is 151° and the size is 152. Orientation 153 and Posture 1
54 etc. are written.

Furthermore, as shown in FIG. 14(a), the tablet 1
Position, size, and orientation of components by 6 etc.

Change posture, etc. (in case of position, Xp Y−+X′r ”j
’) to interactively edit the schematic shape according to the image information you are searching for.

Edit the schematic diagram 18 and change the attributes of the constituent elements (shape size,
When the direction, posture 1 position, etc.) is changed, the attribute value of the corresponding component of the search graphic editing list 14A is immediately changed accordingly, as shown in FIG. 14(b).

In this way, by using the editing function of the drawing processor I3, the user can interactively bring the displayed schematic shape closer to the desired image information he or she has in mind.
At the same time, the attribute information of each component can be changed on the search graphic editing list 14Δ. The finally edited search graphic editing list becomes search information, ie, a key image, for searching for desired image information.

However, a search image attribute list is created in advance for each image information to be stored, and as shown in FIG. (correspondence list for search) is prepared. Then, at the time of searching, the user uses the drawing processor +3 to check the search figure editing list compiled by the user for desired image information and the search image attribute list defined in advance at the time of image information storage, and select each component and each One or more image information having a high degree of similarity with respect to the attribute information of are presented as candidate images.

Unlike keywords, the search figure editing list used as search information in the key image search method has various attribute information (size, direction, posture, position, etc.) of the constituent elements of one image information, so search by keyword It is possible to search for candidate image information that is closer to the image the user has in mind.

However, in the key image search method, a predetermined standard is used as a fixed criterion for determining the degree of similarity for each attribute information of the constituent elements in the user-created key search graphic editing list.

For example, the schematic diagram 18 presented by the user is shown in FIG.
), if the image shows the apple 7 in the center of the desk 6, is it really that Δ from the center of the desk 6 in the same figure (b)?
Should an image in which apple 7 is located at a position shifted by a certain amount (indicated by a broken line) be presented as a candidate image?

The judgment was made based on a predetermined threshold value T. For this reason, it has the disadvantage that it is very difficult to flexibly adapt to the user's search requests regarding the layout.

(Objective of the Invention) The object of the present invention is to allow the user to semi-automatically create a schematic shape as a search key, and to use this to recognize and recognize the schematic shape.
In a key image search method for images that efficiently searches for similar image information by omitting the process of understanding, the user himself/herself judges the similarity jAi? for each individual search. The purpose is to allow I to be easily set.

(Structure of the Invention) (Characteristics of the Invention and Differences from the Prior Art) In the present invention, a key image search method uses a language input format drawing/editing processor to create a schematic shape for searching image information desired by the user. The main feature of this method is that it allows the user to specify criteria for determining similarity when creating a key image. Being able to specify this similarity criterion means that when creating a key image as shown in Figure 16(a), for example, various attribute information (position, size, ,
This means that the user needs to specify the judgment criteria for each position (posture 1, front and back, etc.).

This is a very troublesome task for the user.

Therefore, the second feature of the present invention is that the similarity criterion is automatically estimated from the editing information when the user creates and edits the key image using the drawing processor [3]. Therefore, in the present invention, it is possible to specify the criteria for determining the degree of similarity without increasing the burden on the user at all, and this point is different from the conventional method.

That is, in the conventional key image search method, the similarity criterion is predetermined for each attribute information, and the user cannot directly specify this criterion. Furthermore, even if it were possible to directly specify the criteria, it would be very time-consuming to specify the criteria for each attribute of each component when creating a key image.

(Embodiment) FIG. 1 is a configuration block diagram of an embodiment of the present invention,
21 is an input section, and 22 is an image component calling section.

23 is an image component editing section, 24 is a display section, 25 is a candidate image information evaluation section, and 26 is a search correspondence list storage section.

27 is an image information search control section, and 28 is an image information storage section.

29 is an editing information input section, and 30 is a similarity criterion estimation section.

However, the image information storage section 28 has stored T images of image information in advance, and the search correspondence list storage section 26 stores information about each of the T images stored in the image information storage section 28. As shown in Figure 2, each component name of the image information 14. The relationship between the default attribute information 15 and the storage position 2 in the image information storage unit 28 and the correspondence list 4 is stored in a table format, and identification symbols 3 and 3 are provided for each piece of image information in order to select candidate images, which will be described later. is given.

An outline of the operation of this embodiment will be described with reference to FIG. First, when the user inputs the component name 14 of desired image information from the input section 21, the one-image component calling section 22 stores graphic data regarding a plurality of types of schematic shapes and a default graphic editing list for search. The approximate figure data corresponding to the element name and the default search figure aggregation list are output to the image component editing section 23.

When the image component editing unit 23 reads the approximate figure data, it visualizes the approximate figure data and displays it on the display unit 24.

The user inputs editing information from the editing information input section 29 while viewing this. Here, the editing information is information for moving the position or changing the size and orientation of the approximate figure being displayed on the display unit 24, and is input from a tablet or the like.

When the image component editing unit 23 reads the editing information, it moves the position of the schematic figure being displayed on the display unit 24, changes its size and orientation, and also edits the schematic figure that is being displayed. At the same time, the search figure edit list is also changed. In this way, the user edits each component displayed on the display unit 24 so that the position, size, orientation, etc. of each component displayed on the display unit 24 match the corresponding component of the desired image information.

Further, when the similarity criterion estimating unit 30 reads the editing information from the image component editing unit 23, the similarity criterion estimating unit 30 estimates the similarity criterion of attribute information such as the position, size, and posture of the approximate figure, and The similarity determination reference value is output to the search figure editing list, and the reference value is displayed on the display unit 2 as a rough figure.
Output to 4.

The candidate image information evaluation unit 25 reads out the search graphic editing list created by the image component editing unit 23, and searches for similarities between the constituent elements of the search graphic editing list and their attribute information from the search correspondence list storage unit 26. Degree judgment JAI? ! The storage position of one or more pieces of image information within the value is read out and outputted to the image information search control section 27. The image information search control section 27 reads out the image information corresponding to the storage position of the image information storage section 28 and displays it on the 1 display section 24.

FIG. 3 shows the image component calling section 222, image component editing section 23, and the like shown in FIG. A detailed block diagram of the candidate image information evaluation unit 25 and the similarity criterion estimation unit 30 is shown.

Each of these parts will be explained below.

The image component calling unit 22 is an image component search control unit 22.
1 and an image component storage section 222.

As shown in FIG. 4, this image component storage unit 222 includes symbols representing various components and graphic data 155 for displaying the components as approximate figures. Further, a default value default attribute value vd15 of attribute information is stored as a search graphic editing list of the component.

In this embodiment, a symbol representing a component is given a specific name of the component (such as "apple" or "desk") and is referred to as a component name.

In addition, as shown in Figure 5, the figure data for displaying the approximate figure is a two-character set array (x
o, yO)(xl, yl), . When expressing a schematic figure as a solid in a three-dimensional space, one figure data is a three-character set (xo, yO, ZO) (XI, yl, zl), which represents the vertex coordinates in the three-dimensional space.
, and phase information representing the connection of vertex coordinates.

Further, here, the constituent elements are assumed to be schematic figures in a two-dimensional space, the degree of the attribute is set to N=3, the -secondary attribute is set to "position", the secondary attribute is set to "posture", and the tertiary attribute is set to "r size".

- The next attribute "position" is the center of gravity of the rectangle circumscribing the schematic shape of the component, and its coordinates are the two-character set (xc, yc)
Expressed as

The secondary attribute "posture" is expressed as a unit vector p in a two-dimensional space whose starting point is "position", and in the default state, p of each component is the initial position vector pi (pi is a certain unit vector in the two-dimensional space). shall be equal to
In accordance with the rotation of the schematic figure, p is also rotated by the rotation angle.

The tertiary attribute "size" represents the enlargement/reduction magnification of the approximate figure, and the default value for each component is 1.

When treating a schematic figure as a solid in three-dimensional space,
- The second attribute may be a unit vector in the three-dimensional space, with the three-letter set (XC, yc, zc) representing the coordinates of the center of gravity of a rectangular parallelepiped circumscribing the approximate figure.

The image component editing unit 23 is a drawing/editing processor 231
and a search graphic editing list storage section 232.

This drawing/editing processor 231 has the same functions as commercially available drawing/editing processors such as personal computer software that generates and displays figures composed of straight lines, arcs, etc., and edits their position, orientation, size, etc. have.

In addition, the 1 drawing/M collection processor 231 reads out the graphic data of the constituent elements stored in the search graphic list storage unit 232, and displays a figure in which two character sets are connected clockwise with a straight line on the display unit 231.
Display on 4.

If the figure data is a solid in a three-dimensional space, the three-dimensional coordinates indicated by the three-character set of one figure data are converted into two-dimensional coordinates by perspective projection transformation, and the converted vertex coordinates are arranged in a straight line in the order indicated by the phase information. The connected figures are displayed on the display section 24.

The editing information input unit 29 is a commercially available tablet or the like, and the user inputs editing information from the editing information input unit 29 to the display unit 24 to move the position, change the posture, or change the size of the figure being displayed. When the drawing/GR collection processor 231 reads the editing information from the editing information input section 29, it moves the position of the figure being displayed on the display section 24, changes its posture, or changes its size according to the editing information. The two-candidate image information evaluation unit 25 that changes the image quality is the preliminary candidate storage unit 2.
51° component comparison section 252. Basic candidate storage unit 253.
-Next attribute evaluation section 254. - Next candidate storage section 255. Secondary attribute evaluation section 256. Secondary candidate storage unit 257. Tertiary attribute evaluation unit 258 and tertiary candidates? It is composed of nine stacking parts 259. The functions of each part of this item will be described later when explaining the operation.

The secondary and tertiary attribute evaluation units 254.256.258 and the candidate storage unit 255.257.259 are at the above-mentioned “positions”.

Used to evaluate and accumulate "posture" and "size".

The similarity criterion estimating unit 30 is a -next editing information accumulating unit 30
1. -order similarity criterion estimation unit 302. Secondary editing information storage section 303. Secondary similarity criterion estimation unit 304; Tertiary editing information storage unit 305. and a tertiary similarity criterion estimation unit 306.

The similarity criterion estimating unit 30 receives the values of each attribute information edited by the image component editing unit 23, and then stores the values in the corresponding edited information storage units 301, 303, and 305. For example, if the position is a primary attribute, when the position of the component is edited in the 1 and 2 image component editing section 23, the ki of this editing is
The history is accumulated in the primary editing information storage unit 301 for each editing step. However, the edited information that is accumulated is only the one step before the current editing step, and the edited information before that is discarded.

For example, when the position of a component is moved as shown in FIG.
Next edit information? Based on the editing information accumulated in the S product unit 301, the primary similarity determination base and estimation unit 302 estimates the similarity determination base and value of this attribute.

Is this estimation stored in the primary editing information a storage unit 301? I multiplied J
It is determined as the standard deviation value of the attribute value of the 41st collection information.

For example, since the position of the -th attribute is secondary paper (xc, yc), the -th edition information storage unit 301 stores the secondary paper (xci, yci) i=0. ,,,t-1
is accumulated. Therefore, the similarity criterion value is (xci
.

yci) i=o, , , , Let the standard deviation of t-1 be S + Sy.

Xm1n: xc -5x YIlin = yc -sy X+*ax = xc + sxY■ax
= yc + ay [(Xmin, Ym
in), (Xrxax, Ytzax) as vertices.

This similarity determination reference value is outputted to the image component editing section 23 and outputted as a graphic to the display section 24 as illustrated in FIG. 7. With this configuration, the user can explicitly observe what kind of similarity criterion value is estimated by his/her editing operations.

Next, the operation of this embodiment will be explained with reference to FIG. The input unit 21 inputs a symbol representing a component of the desired image information. In this embodiment, the input symbol is a component class name that specifically indicates what the component is. Enter ``desk'' etc.

The image component search control unit 221 reads the component name input to the input unit 21 and searches the image component storage unit 222 for a component name that is the same as the component name.

If so, the image component search control section 221 reads the graphic data and attribute information of the component name from the image component storage section 222 and stores them in the search graphic editing list storage section 232. Although the search is performed using a search method that has been used previously, since each component of the image information is searched for, identification is easy.

The search graphic editing list storage unit 232 stores the component name 80, graphic data 81 . Attribute information 821,
822.823, and similarity determination reference values 8211.8221.8231 for each attribute information are accumulated as shown in the figure. The drawing/editing processor 231 reads the component graphic data 8I stored in the search graphic list storage section 232, visualizes and displays the graphic data on the display section 24.

All of the constituent elements included in the desired image information are displayed on the display section 24.
The above procedure is repeated until the component name 80° figure data 81. is displayed in the search figure edit list storage section 232. Add attribute information 82L 822.823.

Next, the user inputs editing information from the editing information input section 29.
When editing information is read from 9, the figure displayed on the display section 24 is moved according to the editing information.

At the same time as the rotation, enlargement/reduction, the attribute value of the component corresponding to the figure in the 11 search figure collection list storage section 232 is also rewritten.

Furthermore, the attribute values changed as a result of the editing are determined by the similarity judgment Jk and the corresponding editing information storage units 301 and 3 of the estimation unit 30.
At 03.305? Multiply S. This situation is shown in FIG.

The position of the figure (apple) displayed on the display section 24 is (
When changing from (xC, yc) to (s+ye'), the primary attribute corresponding to the shape in the search graphic editing list storage section 232 changes from (xc, yc) to Cxc r
yc') to IJ. At the same time, a new similarity judgment value is updated, and the attribute value (xc' v
yc') is - next editing information? The data is stored in the 7flx section 301.

Also, if we rotate the figure to change its posture and change the vector p representing the posture of the figure from pi to pe,
In conjunction with this, a new similarity judgment value is updated, the secondary attribute corresponding to the shape in the search graphic edit list storage section 232 is rewritten from pi to pe, and the secondary attribute is stored in the secondary edit information storage section 303. Attribute value pe is accumulated.

When the size of a figure is enlarged by 1.3 times, the tertiary attribute corresponding to the figure is rewritten from 1 to 1.3.

The new ri similarity determination value is updated, and the tertiary attribute value 1.3 is stored in the tertiary editing information storage unit 305. In this way, the position of each component of the search graphic editing list storage section 232,
Editing is performed by changing the posture and size so as to approximate the position, posture, and size of the component included in the desired image information.

Each edit information storage unit (301, 303, 305) is capable of storing attribute values for t steps, and if it becomes necessary to store attribute values exceeding t steps while one edit progresses, the most Old attribute values are discarded and the latest attribute values are stored in their place.

The -order similarity criterion estimating unit 302 reads out all attribute values stored in the primary editing information storage unit 301.

The zero drawing/editing processor 231 obtains the similarity judgment reference value at that point and outputs it to the search figure editing stroke M product unit 232. This is read out, converted into a graphic, and outputted to the display unit 24.

Similarly, the second-order and third-order similarity criterion estimation unit: 10
4.306 is the secondary and tertiary editing information storage section 303゜3
Read all attribute values multiplied by i in 05.

The similarity determination reference value at that time is determined and outputted to the search graphic editing list ¥1 product unit 232, converted into a graphic in the same manner as described above, and outputted to the display unit 24.

When one edit is completed, the component comparison unit 252 reads out the first component name stored in the search graphic edit list storage unit 232, and then includes the first component name from the search correspondence list storage unit 26. The identification numbers of all image information are read out and stored in the basic candidate storage section 253 and preliminary candidate storage section 251.

Next, the component comparison unit 252 reads out the second component name stored in the search graphic editing stock exchange unit 232, and then sequentially reads out each identification number stored in the preliminary candidate storage unit 251. , it is checked whether the component name is included in the search image attribute list of the image information corresponding to the identification number in the search correspondence list storage unit 26. If not included, is it a backup candidate? The identification number stored in the 9-fold section 251 is deleted, and if it is included, it is left as is.

Next, the component comparison section 252 reads the contents of the preliminary candidate storage section 251 and writes them into the basic candidate storage section 253.

This is repeated, and if all the identification numbers in the preliminary candidate storage section 251 are deleted, the component comparison section 252 stops its operation, and the -next attribute evaluation section 254 starts its operation. (The operation of the -next attribute evaluation unit 254 will be described later.) The component comparison unit 252 configured in this manner stores a search correspondence list for all the component names stored in the search graphic edit list storage unit 232. It is searched whether there is image information having the component name in the section 26.

With this configuration, when the search for all component names is completed, the contents of the basic candidate storage section 253 will be the image containing the most component names stored in the search graphic editing list storage section 232. The information identification numbers are multiplied by M.

The primary attribute evaluation unit 254 is the search graphic editing list storage unit 2
The primary attributes of the first component and their similarity judgment criteria stored in the basic candidate storage unit 32 are read out, and then the basic candidate storage unit 2
53, and reads out the primary attribute of the component having the same name as the component from the search image attribute list of the image information with the identification number in the search correspondence list storage section 26. .

Then, it is determined whether the primary attribute value of the subsequent line is within the former similarity determination reference value. If it is not within the similarity determination j549 value, the evaluation of the first candidate is stopped and the evaluation of the second candidate is immediately started. If the similarity is within the similarity criterion value, the primary attribute evaluation unit 254 similarly evaluates the search image attributes of the image information for the primary attributes of all the constituent elements stored in the search graphic editing list storage unit 232. Evaluation is performed using the primary attribute of the corresponding component of the list and the similarity criterion value.

When the comparative evaluation of all components was completed, the image information of the first candidate had not yet been rejected.
- The identification number of the image information is stored in the next candidate ¥1 accumulation unit 255.

Next, the primary attribute evaluation unit 254 evaluates the primary attributes of all the constituent elements of the second candidate image information in the basic candidate storage unit 253 in the search graphic editing list storage unit 232 in the same manner as in the case of the first candidate image information. Evaluate using the primary attribute of each component and its similarity criterion value.

If the similarity is within the similarity determination reference value, the identification number and evaluation value of the image information of the second candidate are stored in the primary candidate storage section 255.

In this way, the primary attribute evaluation unit 254
The degree of compatibility with the primary attributes of the components stored in the search graphic edit list storage unit 232 is evaluated for all of the 53 image information, and the primary attribute values of all the components are within the similarity determination reference value. Only candidate images are used as primary candidates'a product part 2
Accumulates to 55.

Similarly, the secondary attribute evaluation section 256
It is determined whether the secondary attribute value of each component of all the image information in step 5 is within the similarity determination reference value of the secondary attribute of the corresponding component stored in the search graphic edit list storage unit 232. . If the secondary attribute values for all components are within the similarity criterion value, the secondary candidate storage unit 2
The identification number of the image information is stored in 57.

Similarly, the tertiary attribute evaluation unit 258 calculates the tertiary attribute value of each component of all the image information in the secondary candidate storage unit 257 for the corresponding configuration stored in the search graphic editing list ¥1 product unit 232. Tertiary attribute value of the element.

6. If the tertiary attribute values of all the constituent elements are within the similarity determination reference value, the identification number of the image information is stored in the tertiary candidate storage unit 259.

The image information search control unit 27 sends 79
Read out the identification number of the M candidate image.

The image information corresponding to the identification number is read from the image information storage section 28 and displayed on the display section 24.

(Effects of the Invention) As explained above, the present invention provides a key image search method in which a schematic diagram for searching for desired image information is created using a drawing/editing processor in a language input format.
Users can specify similarity criteria when creating a key image. In particular, the present invention requires specification of similarity criteria.

Since it is automatically estimated based on the history of key image editing operations, the user is freed from the trouble of specifying the similarity criterion value each time. That is, in the present invention,
It becomes possible to specify the similarity determination reference value without increasing the user's burden at all in specifying the similarity determination reference value.

[Brief explanation of the drawing]

FIG. 1 is a configuration block diagram of an embodiment of the present invention, and FIG. 2 is a list of correspondence between graphic data and attribute information of each component name of the image information stored in the image information storage unit 28 of FIG. 1. 3 is a diagram showing the detailed block configuration of the main part of FIG. 1, and FIG. 4 is a diagram showing the graphic data and attribute information ( position,
5 is a diagram for explaining the vertex coordinates of a schematic figure, and the fjSB diagram is a diagram for explaining the storage operation in the primary editing information storage unit 301 due to movement of the position of the component, Figure 8 shows an example of the names of constituent elements stored in the search graphic editing list storage unit 232, FIG. 9 shows the primary, secondary, and tertiary editing information storage unit 30.
1.303.305 is a diagram showing an example of component names, etc. and attribute values, Figures 10 to 12 are explanatory diagrams of each side searching for desired image information from a conventional image database, and Figures 13 and 12 are FIG. 14 is an explanatory diagram using the drawing processor 13 in the language input format using the conventional key image search method.
Figure 15 shows the storage location and search l! of each image information used in the language input format mentioned above. # An example of the correspondence table of the image attribute list, No. 16 [4] is a diagram for explaining the correspondence between the schematic figure 18 and the image information 17 according to the conventional key image search method. 21...Human power department, 22...Image component calling unit, 2
3... Image component editing section, 24... Display section, 25... Candidate image information evaluation section, 2
6... Correspondence list for search TI part. 27... Image information search control unit, 28... Image information storage unit, 29... Edit information input unit, 30... Similarity criterion estimation unit, 221... Image component search control unit, 222...Both image component storage section, 231...Drawing/editing processor, 232...Search m figure [collection list storage section, 251...Preliminary candidate storage section. 252...Component comparison section, 253...Basic candidate storage section, 254...Next attribute evaluation section. 255...-Next candidate? ? f & section, 256... Secondary attribute evaluation section, 257... Tertiary candidate storage section, 258...
Tertiary attribute evaluation unit, 259... Tertiary candidate storage unit, 301
...-Next editing information storage section. 303...Secondary editing information \i11 part, 305...Tertiary editing information storage unit, 302...-order similarity criterion estimating unit, 304...secondary similarity criterion estimating unit, 306
...Tertiary similarity criterion estimator. Patent Applicant Nippon Telegraph and Telephone Corporation Figure Figure 15 Default Figure Figure Figure Figure 7...Ishikio +4A...Aki Sokei 2 Colors Birds Collection List 5... Te"

Claims (1)

[Claims] an image component calling unit that stores components such as graphical information representative of individual materials constituting image information and their attribute information in advance and calls a desired component in response to a user's request; and the image configuration. An image component editing section that edits attribute information such as the position, size, shape, and posture of each image component called by the element calling section, and similarity determination criteria based on the user's editing information in the image component editing section. a similarity determination reference value estimator for estimating a value, a storage position of image information stored in a storage medium such as a magnetic disk, and a position in the image information for decomposing the content of the image information into its constituent elements; A search correspondence list storage unit that defines attribute information such as size, shape, posture, etc., attribute values of constituent elements of each image in the search correspondence list, and attributes of constituent elements edited by the image constituent element editing unit. and a candidate image information evaluation unit that compares and evaluates the image information with the estimated similarity judgment base value and selects image information having a high degree of similarity between the two as a search candidate. Information retrieval device.