JP2001117928A - Device and method for retrieving similar image and recording medium recording similar image retrieving processing program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a device and a method for retrieving similar image and a recording medium recording similar image retrieving processing program, with which the intention of a retriever can be directly fetched. SOLUTION: Concerning the similar image retrieving device for retrieving an image similar to a reference image to become the reference or retrieval from an image database storing plural images, this similar image retrieving device is provided with a means for designating an area on the reference image, a means for designating information expressing the intention of retrieval to the area, and a means for retrieving the image similar to the reference image from the image database while considering the retrieval intention.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

[0001] 1. Field of the Invention [0002] The present invention relates to a similar image search device, a similar image search method, and a recording medium on which a similar image search processing program is recorded, which can directly capture the intention of a searcher.

[0002]

2. Description of the Related Art As examples of conventional similar image retrieval apparatuses, for example, Japanese Patent Application Laid-Open Nos. Hei 7-21198 and Hei 8-249.
352 publication.

[0003] In Japanese Patent Application Laid-Open No. 7-21198,
It has been proposed to select a plurality of images similar and dissimilar to the desired image as reference images and perform an image search using all the feature amounts of the reference images.

Japanese Patent Application Laid-Open No. 8-249352 proposes a method in which a plurality of similar images and dissimilar images are designated, and a reference position for searching in a feature space is changed using the information. I have.

Hereinafter, such a conventional example will be described in some detail.

FIG. 7 is a block diagram showing a configuration of an example of such a conventional similar image search apparatus.

[0007] In FIG. 7, reference numeral 1001 denotes a similar image to an image similar to an image desired by a user and an image not similar to the desired image among a plurality of images displayed on the screen. Similar image / non-similar image designation as a dissimilar image.

[0008] Reference numeral 1002 denotes a feature amount extraction unit for extracting feature amounts quantitatively representing features such as colors and shapes of a designated image.

Reference numeral 1003 denotes a feature vector creation unit that creates a vector having the feature amount extracted by the feature amount extraction unit 1002 as a constituent element.

Reference numeral 1004 denotes a similarity search engine for calculating the similarity between the feature vector created by the feature vector creation unit 1003 and each vector in a feature vector database 1006 described later.

An image database 1005 stores a plurality of images as a database.

Reference numeral 1006 denotes a feature vector database in which feature vectors of images in the image database 1005 are stored as a database.

Reference numeral 1007 denotes a search result display section for displaying an image having a high degree of similarity calculated by the similar search engine 1004 as a search result on a screen.

Next, the operation will be described.

FIG. 8 is a flowchart showing the operation of the conventional similar image search apparatus.

First, a similar image / non-similar image designating section 100
In step 1, an image similar to the image desired by the user or a dissimilar image is designated from a plurality of images on the screen (ST
1001).

Next, in feature value extracting section 1002, feature values quantitatively representing the features of the designated image are extracted from each designated image (ST1002).

Subsequently, the feature vector creation unit 1003 creates a feature vector serving as a search reference position in the feature space using the feature amount of the similar image and the feature amount of the dissimilar image (ST1003).

Further, similarity search engine 1004 calculates the similarity between each vector in feature vector database 1006 and the feature vector serving as a search reference position (ST1004).

Next, in the search result display section 1007,
The images in the image database 1005 are displayed in the order of high similarity calculated by the similar search engine 1004 (S
T1005).

Then, a series of procedures up to this point are repeated until an image satisfying the search intention is obtained, and the process ends when the search intention is satisfied (ST1006).

In the above, an example in which a plurality of similar images and non-similar images are designated to perform similar image retrieval has been described. Instead of designating an image, a method of designating an object or an area in the image has been proposed. I have.

Japanese Patent Laying-Open No. 10-154149 proposes specifying an object similar to a desired object and recalculating the feature amount weight.

In Japanese Patent Application Laid-Open No. 11-39332, a favorite feature amount and a dislike feature amount are designated for each area, and a positive weight is assigned to the favorite feature amount and a negative weight is assigned to the feature amount disliked. It is proposed to give the weight of.

The main point of these proposals relates to a method of appropriately updating the weight between the components of the feature vector.

[0026]

However, since the conventional similar image retrieval apparatus is configured as described above, even if an image close to a desired image is presented,
There has been a problem that the user's intention to check whether there is an image in which a specific portion has been improved cannot be directly captured.

Incidentally, even in the conventional similar image retrieval apparatus,
Even in a method of designating a region of interest and selecting an image close to the region, if a desired partial region is available, it may be possible to express it as a combination thereof.

However, there is no guarantee that the combination exists, and continuing to search until they appear on the screen is rather distant.

The present invention has been made to address these circumstances, and has as its object to determine how an image similar to the currently displayed image has an improved specific portion. A similar image search device and a similar image search method capable of directly capturing a searcher's intention that can satisfy a search request of a user who wants to check, and a recording medium storing a similar image search processing program. To provide.

[0030]

According to the present invention, in order to solve the above-mentioned problems, (1) a similar image for searching an image similar to a reference image serving as a search reference from an image database storing a plurality of images; Means for specifying a region on the reference image, means for specifying information expressing a search intention for the region, and an image similar to the reference image in consideration of the search intention. And a means for searching from a database.

Further, according to the present invention, there is provided a method for solving the above-mentioned problem, (2) a method of searching a similar image similar to a reference image serving as a search reference from an image database storing a plurality of images. A step of specifying an area on the reference image; a step of specifying information expressing a search intention for the area; and a step of searching for an image similar to the reference image from the image database in consideration of the search intention. , A similar image search method is provided.

According to the present invention, in order to solve the above-mentioned problems, (3) a processing program for retrieving a similar image similar to a reference image serving as a reference for retrieval from an image database storing a plurality of images by a computer. Is recorded on the recording medium, wherein the processing causes the computer to specify the area on the reference image to the user, the user to specify the information expressing the search intention for the area, taking into account the search intention According to another aspect of the present invention, there is provided a recording medium storing a similar image search processing program for causing an image similar to the reference image to be searched from the image database.

[0033]

Embodiments of the present invention will be described below with reference to the drawings.

(First Embodiment) A first embodiment of the present invention will be described with reference to FIG.

FIG. 1 is a block diagram showing the configuration of the similar image retrieval apparatus according to the first embodiment.

In FIG. 1, reference numeral 101 denotes a reference image selection unit for selecting one image from several display images as a reference image for similar image search.

Reference numeral 102 denotes a search insufficiency area specifying unit for specifying an area in which the user's search intent is not satisfied.

Reference numeral 103 denotes a target setting unit for setting the characteristics of the search unsatisfied area to the value of the search intention of the user.

Reference numeral 104 denotes a feature amount extraction unit that extracts a feature amount that quantitatively represents a feature related to the reference image and the unsatisfied search intention area.

Reference numeral 105 denotes a reference image feature vector creating unit for creating a feature vector having the feature amount of the reference image extracted by the feature amount extracting unit 104 as a constituent element.

Reference numeral 106 denotes a search intent for creating a target feature vector from the feature amount of the unsatisfied search intent area extracted by the feature amount extraction unit 104 and the target value set by the target setting unit 103. This is the target feature vector creation unit for the unsatisfied area.

Reference numeral 107 denotes a combined feature vector creating unit for creating a feature vector combining the target feature vector and the feature vector of the reference image.

Reference numeral 108 denotes a similarity search engine for calculating the similarity between each vector in the later-described feature vector database 109 and the combined feature vector.

Reference numeral 110 is an image database storing a plurality of images as a database.

Reference numeral 109 denotes a feature vector database in which feature vectors of the images in the image database 110 are stored as a database.

Reference numeral 111 denotes a search result display section for displaying an image having a high degree of similarity calculated by the similar search engine 108 as a search result on a screen.

Next, the operation will be described.

FIG. 2 is a flowchart showing the operation of the similar image retrieval apparatus according to the first embodiment of the present invention.

First, at the start of the search, it is determined whether or not there is an image close to the search intention on the screen (ST101).

According to this determination, if no image is displayed on the screen, or if an image completely different from the image desired by the user is displayed, the target setting unit 103 For a feature that can be understood by the user, such as a color, a target (desired value) is selected from, for example, a menu-type screen and a search is requested (ST10).
2).

Next, in response to this search request, the target feature vector creating unit 106 for the unsatisfied search intent area calculates a target feature vector from the value of the selected feature amount (ST103).

Of course, at this time, the search unsatisfied area can be regarded as coincident with the entire image.

Next, the degree of similarity between the target feature vector thus created and each vector in the feature vector database 109 is calculated (ST104).

Subsequently, the images in the image database 110 are presented to the user as a search result in the descending order of the calculated similarity value (ST113).

Here, it is determined whether or not there is an image satisfying the search intention on the screen (ST114).

At this time, if an image satisfying the user's search intention is obtained, the search is terminated. If not, the process returns to step ST101.

Here, if no image is displayed at the start of the search or if an image completely different from the image desired by the user is displayed, targets (desired values) relating to some features are set. Although an example has been described in which a search is performed based on a combination of designations, a keyword search or a search using a simple sketch may be used instead.

Now, in step ST101, when there is an image close to the search intention on the screen,
The operation when searching for an image similar to the image displayed on the screen will be described in detail below.

First, the reference image selecting section 101 designates the image as a reference image (ST105).

Next, search unsatisfied area specifying section 102 specifies a search unsatisfied area (ST10).
6).

At this time, the area is specified by, for example, using a pointing device such as a mouse to surround the area in a rectangular or circular shape.

Subsequently, the target setting unit 103 sets the feature of the unsatisfied search intention area to the value of the user's search intention, and requests a search (ST107).

Here, the target setting section 103 may specify not only specific values but also a value range, a magnitude relationship with the current value, and the like.

For example, the user selects one of options prepared in advance in a menu-type screen.

As a specific example, when there is a face image with an open mouth and a face image with a closed mouth is desired, a region of the mouth is selected, and a horizontal bar close to the shape of the closed mouth is selected.

When there is a face image in which eyes are closed and a face image in which the eyes are open is desired, an eye area is selected.
For example, select a circle close to the shape with the eyes open.

If the user cannot express a specific request well, the target is selected to be "no particular".

This case is dealt with, for example, by retrieving a plurality of feature spaces separated by a certain distance or more in the feature space.

The relationship between the change request regarding colors and shapes that can be used as a range of common sense of general users and the values of feature quantities such as hue, saturation, lightness, circularity, complexity, and fractal dimension is as follows. It is set in advance, or the association is made adaptively by learning.

Next, the feature quantity extracting unit 104 extracts the feature quantity of the reference image and the feature quantity of the unsatisfied search intent area (ST108).

Subsequently, the target feature vector creation unit 106 for the search unsatisfied area creates a target feature vector for the search unsatisfied area (ST10).
9).

Here, an example of a method for creating a target feature vector will be briefly described.

First, it is assumed that a feature vector having a feature amount of a search unsatisfied area as a component is a.

When a turret is designated and is represented by a feature value associated in advance, a target vector of a search unsatisfied area is obtained by replacing the corresponding component of the vector a with the value of this feature value. .

Subsequently, the feature vector creating section 1 for the reference image
At 05, a feature vector of the reference image is created (S
T110).

Next, the combined feature vector creating unit 107 creates a combined feature vector obtained by combining the target feature vector and the feature vector of the reference image (ST11).
1).

Next, in the similar search engine 108,
The similarity between each vector in the feature vector database 109 and the combined feature vector is calculated (ST112).

Subsequently, all vector data whose calculated similarity is larger than a preset threshold are selected.

Alternatively, sorting may be performed in the order of larger similarity values, and some higher-ranked items may be selected.

An image corresponding to the vector data thus selected is extracted from the image database 110 and displayed on the search result display unit 111 (ST113).

If there is an image satisfying the user's search intention among the search results (ST114), the process ends.

However, if the user's search intention is not yet satisfied, the process returns to step ST101 to select an image closest to the user's search intention from the images being presented (ST105). Repeat a series of procedures up to.

The method of designating a search unsatisfied area and the like is not limited to a method of enclosing the area with a pointing device, for example, a method of specifying a position using voice recognition, a method of using an interface such as a data glove, or the like. And so on.

(Second Embodiment) FIG. 3 is a block diagram showing the configuration of a similar image retrieval apparatus according to a second embodiment of the present invention.

Although the configuration of the second embodiment is substantially similar to the configuration of the first embodiment, the main difference from the first embodiment is that the search intention of the user is almost satisfied. Search-intention-satisfied-region specifying unit 2 for specifying a region in a reference image
The second embodiment is different from the first embodiment in that the feature vector creation unit 105 for the satisfaction area of the search intention is provided instead of the feature vector creation unit 105 for the reference image, which is included in the first embodiment.

Therefore, the reference image selecting unit 201, the search insufficiency area specifying unit 203, the target setting unit 20
4. Regarding the feature amount extraction unit 205, the target feature vector creation unit 207 of the unsatisfied search intention area, the composite feature vector creation unit 208, the similar search engine 209, the image database 211, the feature vector database 210, and the search result display unit 212 Are the same as the corresponding contents described in the first embodiment.

Next, the operation will be described.

FIG. 4 is a flowchart showing the operation of the similar image search apparatus according to the second embodiment of the present invention.

The operation of the second embodiment is substantially similar to the operation of the first embodiment, but differs in the following point.

One is that a step (ST207) of designating a search intention satisfaction area is added.

The other is to extract a feature (ST205).
In the first embodiment, the process of creating a feature vector (ST206) is performed on the entire image. However, in the second embodiment, the process is performed on a search intention satisfaction region. .

Accordingly, when creating a combined feature vector (ST208), instead of the feature vector of the reference image,
Satisfaction feature vectors of search intention satisfaction regions are used.

Now, description will be made with reference to a slightly specific example.

For example, suppose that a face photograph with eyes closed is displayed on the screen, and a situation is desired in which it is desired to check whether there is a face photograph with eyes open.

In this case, the eyes are the unsatisfied search intention areas that do not satisfy the searcher's wishes, the faces except the eyes are the search intention unsatisfied areas that substantially satisfy the searcher's wishes, and the background is indifferent. Area.

Here, the non-interest area is an area which may or may not be changed by the search.

Therefore, the indifferent region is a region obtained by excluding the search intention satisfying region and the search intention unsatisfied region from the reference image, and it is not necessary to perform any processing on the indifferent region.

(Third Embodiment) In the first and second embodiments described above, search results are selected and displayed as images. However, in the third embodiment, the search results are not limited to similar images. , Similar areas can be searched and displayed.

The third embodiment will be described with reference to FIG.

FIG. 5 is a block diagram showing a configuration of a similar image search device according to the third embodiment.

In the third embodiment, only the unsatisfied search intent area is targeted. The unsatisfied search intent area designation unit 301, the target setting unit 302, the feature quantity extraction unit 303, the unsatisfied search unsatisfied area is used. The target feature vector creation unit 304, similar image search engine 305, feature vector database 306, image database 307, and search result display unit 309 are the same as the corresponding contents described in the first embodiment.

In addition to these, in the third embodiment, a region cutout unit 308 is further provided.

Next, the operation will be described with reference to FIG.

FIG. 6 is a flowchart showing the operation of the similar image retrieval apparatus according to the third embodiment of the present invention.

In the first and second embodiments, the reference image is selected. In the third embodiment, the area is directly specified without selecting the reference image.

First, although close to the user's search intention,
If there is a search intention unsatisfied area where the search intention is not completely satisfied (ST301), the area is designated (ST301).
304).

Next, a target in the unsatisfied search intent area is designated, and a search is requested (ST305).

Subsequently, the feature amount of the search unsatisfied area is extracted (ST306).

Further, a target feature vector of a search unsatisfied area is created (ST307).

If there is no corresponding area in step ST301, a search is requested by specifying a plurality of features such as colors and shapes (ST302), and a target feature vector is created from those feature amounts (ST303).

The similarity between the target feature vector thus created and each vector in the feature vector database is calculated (ST308).

An image including a similar area is picked up by the similarity search engine 305. After the area extracting unit 308 extracts which area in the image is the similar area,
It is displayed (ST309).

As described above, in the present invention, a search intention unsatisfied area designating section for designating an area that does not satisfy the user's search intention and a desired value relating to the color, shape, direction, etc. of the area are designated. Since the target designating unit is provided, the user's search intention can be directly expressed, and the number of searches can be reduced.

Further, in the present invention, since there is also a satisfactory region of interest designation section that the user's search intention is almost satisfied, a portion that is not desired to be changed to a specific target to be improved and a region of indifferent region that does not matter are specified. There is an effect that it is possible to handle them in a distinguished manner, thereby improving search accuracy.

In the present specification described in the above-described embodiment, in addition to claims 1 to 3 described in the claims, the inventions described below as supplementary notes 1 to 12 are provided. include.

(Supplementary Note 1) A similar image search device that searches for an image similar to a reference image serving as a search reference from an image database storing a plurality of images, and satisfies a user's search intention on the reference image. Means for specifying a search intention satisfying area, and means for specifying a search intention unsatisfactory area that does not satisfy the user's search intention on the reference image,
Means for specifying information expressing a search intention with respect to the search intention unsatisfied area; characteristics of the search intention satisfied area;
Means for searching for an image similar to the reference image from the image database by using a feature of the unsatisfied search intention area in which the search intention is added.

(Supplementary Note 2) A similar image search device for searching an image having an area similar to a specified image area from an image database storing a plurality of images, a means for specifying an area on an image, Means for designating information expressing a search intention; means for searching for an image having an area similar to the area in consideration of the search intention; means for extracting an area similar to the area from the searched image And a similar image search device.

(Supplementary Note 3) This is a method of searching an image database storing a plurality of images for an image similar to a reference image serving as a search reference, and a search satisfying the user's search intention on the reference image. A step of specifying an intention-satisfied area; a step of specifying a search-intention-unsatisfied area that does not satisfy the user's search intention on the reference image; And the characteristics of the search intention satisfaction region,
Searching for an image similar to the reference image from the image database using the feature of the unsatisfied search intention area in which the search intention is added.

(Supplementary Note 4) This is a method for searching an image having an area similar to a specified image area from an image database storing a plurality of images. The method includes the steps of specifying an area on an image and determining a search intention for the area. A step of designating the expressed information; a step of searching for an image having an area similar to the area in consideration of the search intention; and a step of cutting out an area similar to the area from the searched image. A similar image search method characterized by having:

(Supplementary Note 5) A recording medium storing a processing program for searching for an image similar to a reference image serving as a search reference from an image database storing a plurality of images by a computer. Causes the computer to specify a search intention satisfying area that satisfies the user's search intention on the reference image, and specifies a search intention unsatisfied area that does not satisfy the user's search intention on the reference image. Let the user specify the information expressing the search intent for the search insufficiency area, and let the user specify the information. The features of the search insufficiency area and the features of the search insufficiency area in consideration of the search intent are added. A recording medium storing a similar image search processing program, wherein an image similar to the reference image is searched from the image database.

(Supplementary Note 6) A recording medium storing a processing program for searching an image having an area similar to a designated image area from an image database storing a plurality of images by a computer. Let the computer specify the area on the image by the user, let the user specify the information expressing the search intention for the area, let the user take into account the search intention, and search for an image with an area similar to the area, A recording medium on which a similar image search processing program is recorded, wherein an area similar to the area is cut out from the searched image.

(Supplementary Note 7) An image database that stores a plurality of images, a feature amount extracting unit that extracts features of images or regions in the image database as quantitative values, and a feature extracted by the feature amount extracting unit. A feature vector creating unit that creates a feature vector having a quantity as a constituent element, a feature vector database that stores feature vectors created by the feature vector creating unit, and a feature vector in the feature vector database. Search engine that calculates the similarity between the image or region and the feature vector created by the feature vector creation unit, and a search result display that displays, as a search result, an image with a high similarity calculated by the similar search engine. And a plurality of images displayed on the screen in a similar image search device comprising A reference image selecting unit that selects one image as a reference image for search, and a search intention unsatisfied area specification that specifies a search intention unsatisfied area that does not satisfy the user's search intention in the reference image And a target setting unit capable of setting a specific desired pattern or desired value or desired range of the search intention unsatisfied area,
The feature vector creating unit, a feature vector creating unit of the reference image, a target feature vector creating unit of the search unsatisfied area, a combined feature vector creating unit that combines the feature vector and the target feature vector of the reference image, A similar image search device characterized by comprising:

According to this configuration, since the apparatus includes the search intention unsatisfied area designating section and the target setting section, the search desire of the user who wants to check whether or not there is an image having a specific portion is determined. Since it is possible to express directly, there is an effect that the number of searches can be significantly reduced as compared with a conventional similar image search apparatus that indirectly infers a user's search intention.

Further, according to this configuration, since the composite feature vector creating unit is provided and the balance between the entire image and the target of the search unsatisfied area can be considered, the search result indicates that the search unsatisfied area is not satisfied. Has the effect that it is possible to prevent a useless search in which the other parts have been significantly changed in spite of satisfying the user's search intention.

(Supplementary Note 8) The similar image search device further includes a search intention satisfying area designating unit that substantially satisfies the user's search intention, and the feature vector creating unit includes a search intention satisfying area satisfying feature. The resemblance description (7), comprising a vector creation unit, a target feature vector creation unit for a search unsatisfied area, and a combined feature vector creation unit that combines the satisfaction feature vector and the target feature vector. Image search device.

According to this configuration, since the apparatus includes the search intention unsatisfied area specifying unit and the search intention satisfied area specifying unit,
As a result, it becomes possible to handle the three areas of the search intention satisfying area to be fixed, the search intention unsatisfying area to be improved, and the indifferent area of concern, so that the processing can be performed by removing the indifferent area such as the background. There is an effect that search accuracy can be improved.

Further, according to this configuration, since the composite feature vector creating unit is provided and the balance between the target of the unsatisfied search intent area and the unsatisfied search intent area can be considered, the result of the search, Regarding the area, there is an effect that it is possible to prevent a useless search in which the search intention satisfying area is largely changed in spite of satisfying the user's search intention.

(Supplementary Note 9) An image database that stores a plurality of images, a feature amount extraction unit that extracts features of images or regions in the image database as quantitative values, and a feature vector that includes the feature amounts as components. A feature vector creating unit for creating a feature vector, a feature vector database storing the feature vectors, and a similarity search for calculating a similarity between a feature vector in the feature vector database and a feature vector created for a specified image or region. In a similar image search device including an engine and a search result display unit that displays an image having a high calculated similarity as a search result, a search intention that does not satisfy a user's search intention in the reference image is not satisfied. A search-intention-unsatisfied-area specifying unit that specifies an area, and a specific desired pattern or Is a similar image search device, comprising: a target setting unit that can set a desired value or a desired range; and an area cutout unit that cuts out a desired area in the image database.

According to this configuration, since the search intention unsatisfied area designating section, the target designating section, and the area cutout section are provided, the user's search intention can be directly expressed. As compared with the conventional similar image search device that estimates the search intention, the number of searches in the search for the region image (component image) can be greatly reduced.

(Supplementary Note 10) A similar image search device that searches for an image similar to a reference image serving as a search reference from an image database storing a plurality of images, and satisfies a user's search intention on the reference image. Means for specifying a search intention unsatisfied area that is not, means for specifying information expressing a search intention for the search intention unsatisfied area, and characteristics of the search intention unsatisfied area in which the search intention is taken into account, Means for searching the image database for an image similar to the reference image.

(Supplementary Note 11) This is a method of searching an image database storing a plurality of images for an image similar to a reference image serving as a search reference, and a search that does not satisfy a user's search intention on the reference image. A step of designating an unsatisfied intention area; a step of designating information expressing a search intention for the unsatisfied search intention area; and a feature of the unsatisfied search intention area in consideration of the search intention. Searching for an image similar to the above from the image database.

(Supplementary Note 12) A recording medium storing a processing program for retrieving an image similar to a reference image serving as a retrieval reference from an image database storing a plurality of images by a computer. Causes the computer to specify a search intention unsatisfied area that does not satisfy the user's search intention on the reference image, and causes the user to specify information expressing a search intention for the search intention unsatisfied area, A recording medium in which a similar image search processing program is recorded, wherein an image similar to the reference image is searched from the image database using characteristics of the search intention unsatisfied area in which the search intention is added.

[0133]

Therefore, as described above, according to the present invention, it is determined whether or not an image similar to the currently displayed image has an improved specific part. Abstract: Provided is an image search apparatus that can satisfy a user's search request, that is, a similar image search apparatus and a similar image search method that can directly capture the intention of a searcher, and a recording medium that stores a similar image search processing program. can do.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a configuration of a similar image search device according to a first embodiment of the present invention.

FIG. 2 is a flowchart showing the operation of the similar image search device according to the first embodiment of the present invention.

FIG. 3 is a block diagram showing a configuration of a similar image search device according to a second embodiment of the present invention.

FIG. 4 is a flowchart showing the operation of a similar image search device according to a second embodiment of the present invention.

FIG. 5 is a block diagram showing a configuration of a similar image search device according to a third embodiment of the present invention.

FIG. 6 is a flowchart showing an operation of the similar image search device according to the third embodiment of the present invention.

FIG. 7 is a block diagram showing a configuration of a conventional similar image search device.

FIG. 8 is a flowchart showing the operation of the conventional similar image search device.

[Explanation of symbols]

101, 201: Reference image selection unit, 102, 203, 301: Search intent unsatisfied area designating unit, 103, 204, 302: Target setting unit, 104, 205, 303: Feature extraction unit, 105: Features of reference image Vector creating units, 106, 207, 304 ... target feature vector creating units 107, 208 ... synthetic feature vector creating units for unsatisfied search intent areas, 108, 209, 305 ... similar search engines, 110, 211, 307 ... Image database, 109, 210, 306: feature vector database, 111, 212, 309: search result display section, 202: search intention satisfaction area designation section, 206: search intention satisfaction area satisfaction feature vector creation section, 308: area extraction Department.

Claims (3)

[Claims] 1. A similar image search apparatus for searching an image database storing a plurality of images for an image similar to a reference image serving as a search reference, means for specifying an area on the reference image, A similar image search device comprising: means for specifying information expressing a search intention; and means for searching the image database for an image similar to the reference image in consideration of the search intention. 2. A method for searching a similar image similar to a reference image serving as a search reference from an image database storing a plurality of images, comprising: specifying an area on the reference image; A method of searching for a similar image, comprising: a step of designating information expressing the above; and a step of searching the image database for an image similar to the reference image in consideration of the search intention. 3. A recording medium storing a processing program for searching a similar image similar to a reference image serving as a search reference from an image database storing a plurality of images by a computer, wherein the processing is performed by the computer. Prompting a user to specify an area on an image, allowing the user to specify information expressing a search intention for the area, and allowing the image database to search for an image similar to the reference image in consideration of the search intention. A recording medium storing a similar image search processing program characterized by the following.