KR20110076089A - System and method for selecting representative image in eliminating image redundancy - Google Patents

Abstract

PURPOSE: A system and method for selecting representative image in eliminating image redundancy are provided to expose the representative image to a search result and improve the search efficiency by eliminating duplicated documents. CONSTITUTION: A redundant image extractor(610) extracts the same images. A search object management unit(640) exposes the same images to the search object. A search execution unit(650) performs an image search about an inputted keyword based on the search object. A search result providing part(660) exposes representative image of the same images when displaying the search results.

Description

Representative image selection method and system in image deduplication {SYSTEM AND METHOD FOR SELECTING REPRESENTATIVE IMAGE IN ELIMINATING IMAGE REDUNDANCY}

The present invention relates to a method and a system for selecting a representative image during image deduplication, and more particularly, to a method and a system for selecting a representative image during image deduplication for exposing all duplicated documents to a search target and removing duplicates in a search. .

In an image search in which an image is searched in response to a user's input keyword, a plurality of identical images may exist in the image to be searched. In this case, a plurality of identical images may be searched and exposed in the image search, which may be a factor to deteriorate the quality of the image search.

On the other hand, when a plurality of identical images are removed to expose only the representative image as a search result, if the exposed representative image does not include the keyword input by the user, the user may search even if an image related to the entered keyword is searched. The accuracy of the results can be questioned.

Therefore, when exposing a representative image by removing a duplicate image, a study on a keyword selection method and system for exposing a representative image including a keyword input by a user is required.

According to the present invention, a duplicate image including a keyword input by a user is exposed to a search result by removing a duplicate document at the time of a search without first removing the duplicate image prior to the search request, thereby increasing the convenience of the search result. Provided are an image selection method and system.

The present invention scores each image by assigning term weights calculated for each word in a document including each image, and determines the image having the highest score as the representative image, thereby representing the keyword including the user input. A representative image selection method and system are provided that allow an image to be exposed to search results.

According to an exemplary embodiment of the present invention, a method of selecting a representative image at the time of removing duplicate images may include extracting a plurality of identical images, exposing the plurality of identical images to a search target, based on the search target. The method may include performing an image search for a keyword and exposing a representative image of the plurality of identical images when the search result is exposed according to the image search.

According to an aspect of the invention, the step of extracting the plurality of the same image, comparing the signature (signature) of each image to determine the image identity, the plurality of images determined to be the same as the plurality of groups It may include extracting the same image.

According to an aspect of the present invention, the signature of each image may be a hash value of each image.

According to an aspect of the present disclosure, the exposing the representative image may include selecting a representative image among a plurality of identical images in the search result, and extracting an image excluding the representative image among the plurality of identical images from the search result. Excluding the representative image may be exposed.

According to an aspect of the present invention, calculating term weights for one or more words in each document including the plurality of identical images and scoring the plurality of identical images using the term weights, respectively. The method may further include: exposing the representative image, and may select an image having the highest score among the plurality of identical images as the representative image.

According to an aspect of the present invention, the scoring of the plurality of identical images may include assigning the determined term weight to the one or more words and adding the term weights assigned to each word for each image to add the plurality of identical images. Score an image.

Representative image selection system at the time of image deduplication according to an embodiment of the present invention, duplicate image extraction unit for extracting a plurality of the same image, a search target management unit for exposing the plurality of the same image to the search target, the search target The search result may include a search performing unit that performs an image search on the input keyword and a search result providing unit which exposes a representative image of the plurality of identical images when the search result is exposed according to the image search.

According to an aspect of the invention, the search result providing unit, the representative image selection unit for selecting a representative image of the plurality of the same image in the search results and the image other than the representative image of the plurality of the same image from the search results Excluding it may include an image exposure unit for exposing the representative image.

According to an aspect of the present invention, a term weight calculation unit configured to calculate term weights for one or more words in each document including the plurality of identical images, and the plurality of same images using the term weights. The apparatus may further include an image scoring unit that scores each, and the search result providing unit may select an image having the highest score among the plurality of identical images as the representative image.

According to an embodiment of the present invention, by removing a duplicate document at the time of a search without first removing the duplicate image prior to the search request, the representative image including the keyword input by the user is exposed to the search result, and the search is performed. A representative image selection method and system are provided that enhance the convenience of the results.

According to an embodiment of the present invention, a keyword input by a user is included by assigning a term weight calculated for each word in a document including each image to score each image, and determining the image having the highest score as the representative image. Representative image selection methods and systems are provided that enable exposing representative images to search results.

Hereinafter, with reference to the contents described in the accompanying drawings will be described in detail the embodiments of the present invention. However, the present invention is not limited to or limited by the embodiments. Like reference numerals in the drawings denote like elements.

FIG. 1 is a flowchart illustrating a representative image selection method when an image is duplicated according to an embodiment of the present invention.

Referring to FIG. 1, in step 110, a plurality of identical images may be extracted. That is, a plurality of identical images may be included in a search target image for image search included in a web page, and the like, and a plurality of identical images may be extracted for image search. An example of a plurality of images for image search will be described in more detail below with reference to FIG. 2.

2 is a diagram for describing duplication of a search target image according to an exemplary embodiment of the present invention.

As shown in FIG. 2, a plurality of images 211, 212, 213, 214, 215, 216, 217, 218, 219, and 220 related to 'Hyo Lee' may be present among the images to be searched. . Here, the images related to 'Hyo Lee' may exist in duplicate of the same image on several web pages. In the example of FIG. 2, image 1 211 may have the same images 213 and 220, and image 2 212 may also have the same image 218. In this case, exposing all of the same images to the search results may reduce the quality of the search and may reduce the user's convenience of the search results. Can be. However, when the representative image is selected and exposed to the search result, the quality of the search result may vary depending on which image is selected and exposed as the representative image. For example, in the case of the image 5 215, since the image is the same as the image 9 219, one of the two images 215 and 219 may be selected as the representative image and exposed. In this case, if the image 9 219 is selected as the representative image, the image 9 219 does not include the keyword 'Hyo-ri', and the user does not include 'Hyo-ri' in the keyword even though the image related to Lee-Ho-ri is exposed as a search result. It can be determined that the search is not performed correctly. Therefore, by inputting a search keyword from the user and performing a representative image selection and duplicate image removal at the time of searching, it is possible to always select a representative image including the search keyword.

Meanwhile, a detailed process of extracting a plurality of identical images will be described in more detail below with reference to FIG. 3.

FIG. 3 is a diagram illustrating an example of a process of extracting a plurality of identical images shown in FIG. 1.

Referring to FIG. 3, in step 310, image identity may be determined by comparing signatures of each image. Here, the signature of the image may be a hash value for the image.

In operation 320, a plurality of identical images may be extracted by grouping a plurality of images determined to be identical to each other. That is, images determined to be identical to each other may be grouped for each image by comparing the image signatures. Therefore, when extracting a plurality of identical images, the same overlapping images may be collected into one group.

Referring back to FIG. 1, step 110 includes calculating term weights for one or more words in each document including the plurality of identical images and using the term weights to generate the plurality of identical images. And further scoring each. That is, weights may be determined for each word by calculating the term weights for a plurality of identical image groups, and the determined weights may be assigned to each word in the document including each image. Therefore, when the weights assigned to each word are summed for each image, the ranking of each image may be determined according to the size of the summed score. Here, the ranked images can then be used to select a representative image, which is described in more detail below.

In operation 120, a plurality of identical images may be exposed to a search target. All of the plurality of images may be exposed to a search target to remove duplicates at the time of searching without removing the duplicates before searching.

In operation 130, an image search for the input keyword may be performed based on the search target. That is, all images including the input keyword may be searched, and the search result may include a plurality of duplicated images.

In operation 140, a representative image of the plurality of identical images may be exposed when the search result is exposed according to the image search. When exposing the result of the search to the user, duplication may be removed and only the representative image may be exposed. In this case, the representative image may be an image including a keyword input from the user.

In this case, as an example, a plurality of identical images may be scored using the term weight calculated above to determine the representative image, and the representative image may be determined using the ranking of each scored image. That is, the highest ranking image among the ranking of the images determined according to the determined term weight may be determined as a representative image of the plurality of overlapping images.

As described above, by eliminating duplicates when performing a search after inputting a keyword without performing deduplication of duplicate images before the search, the representative image can always include the entered keywords, Improve user satisfaction.

4 is a flowchart illustrating a representative image selection method at the time of image de-duplication according to another embodiment of the present invention.

Referring to FIG. 4, in step 410, a plurality of identical images may be extracted. That is, a plurality of identical images for performing an image search can be extracted from a web page or the like.

In operation 420, a term weight may be calculated for one or more words in each document including a plurality of identical images, and in operation 430, each of the plurality of identical images may be scored using the term weight.

In operation 440, a plurality of identical images may be exposed to a search target.

In operation 450, an image search may be performed on an input keyword based on a search target. In operation 460, an image having the highest score among a plurality of identical images may be selected as the representative image.

In operation 470, the representative image may be exposed after excluding an image except for the representative image from the plurality of identical images in the search result.

Here, an embodiment of calculating the term weight and selecting the representative image using the term weight will be described in detail with reference to FIG. 5.

FIG. 5 is a diagram for describing a process of selecting a representative image using a term weight according to an embodiment of the present invention.

Referring to FIG. 5, a plurality of identical images 511, 512, and 513 of the image A of Lee Hyo-ri may be grouped into one group 510. In this case, documents including each of the plurality of identical images 511, 512, and 513 may include different words. To determine a representative image, first, a term weight for words included in each document may be calculated. Can be. The determined term weights may be assigned to words included in each image, words assigned scores may be summed for each image, and each image may be ranked according to the summed score. For example, a plurality of identical images 511, 512, 513 for image A may include the words 'Hyo Lee', 'Sexy', 'Jeon Hyun', and 'Pretty', and sequentially 3.5 for each word. Weights of 1.0, 1.0, and 1.0 may be determined, respectively. Accordingly, the determined weights may be assigned to words included in each image to add scores for each image, and the image having the highest score may be determined as the representative image 530.

As described above, the weight of each word included in the duplicate image is determined, and the representative image is determined by ranking the respective images through the determined weight to determine the representative image including the keyword input by the user at all times. A representative image selection method can be provided.

FIG. 6 is a block diagram illustrating a representative image selection system at the time of image de-duplication according to an embodiment of the present invention.

Referring to FIG. 6, the representative image selection system 600 may include a duplicate image extractor 610, a search object manager 640, a search performer 650, and a search result provider 660.

The duplicate image extractor 610 may extract a plurality of identical images from a web page or the like. That is, a plurality of identical images may be extracted from a web page or the like to provide search target images for image search. Here, the duplicate image extractor 610 may include a duplicate determiner 611 and a duplicate classifier 612. The overlap determination unit 611 may compare signatures of images existing on a web page or the like to determine image identity, and the overlap classification unit 612 may determine a plurality of images determined to be the same as a result of the image identification determination. Can be grouped In this case, the image signature may mean a hash value of the image.

The search target manager 640 may expose a plurality of identical images to the search target. That is, when searching for an image, a plurality of identical images extracted through a web page or the like may be exposed to a search target without removing duplicates of a plurality of identical images before searching.

The search performer 650 may perform an image search for the input keyword based on the search target. Therefore, when the input keyword is included in a word in a document including an image, the image may be searched by the search performer 650.

The search result providing unit 660 may expose a representative image of the plurality of identical images when the search result is exposed according to the image search. That is, by performing an image search based on a search target without removing duplicates of a plurality of identical images and selecting a representative image, an image including the input keyword may be selected as a representative image and exposed to a search result. have.

Here, the search result providing unit 660 may exclude the representative image selection unit 661 selecting a representative image from among a plurality of identical images in the search result and an image excluding the representative image among the plurality of identical images from the search result. The image exposure unit 662 exposing the representative image may be included. Therefore, a search result may be generated by removing duplicate identical images, and the input keyword may be always included in the representative image.

The representative image selection system 600 may further include a term weight calculator 620 and an image scorer 630.

The term weight calculator 620 calculates a term weight for one or more words in each document including a plurality of identical images, and the image scoring unit 630 scores each of the plurality of identical images using the term weights. can do. That is, the image scoring unit 630 may assign one or more words to the determined term weights, and add the term weights assigned to each word to each image to score the plurality of identical images. Here, an image having the highest score among the scored images may be selected as a representative image of the plurality of identical images.

As described above, by removing a duplicate image prior to the search request without removing the duplicate image, the representative image including the keyword input by the user is exposed to the search result and the convenience of the search result is improved. You can.

In addition, by assigning a calculated weight for each word in a document including an image, scoring each image, and determining the image having the highest score as the representative image, thereby always selecting the representative image including the keyword entered by the user. Can be.

In addition, the representative image selection method for image de-duplication according to an embodiment of the present invention includes a computer readable medium including program instructions for performing operations implemented by various computers. The computer readable medium may include program instructions, data files, data structures, etc. alone or in combination. The media may be program instructions that are specially designed and constructed for the present invention or may be available to those skilled in the art of computer software. Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks, and magnetic tape, optical media such as CD-ROMs, DVDs, and floppy disks. Magneto-optical media, and hardware devices specifically configured to store and execute program instructions, such as ROM, RAM, flash memory, and the like. Examples of program instructions include not only machine code generated by a compiler, but also high-level language code that can be executed by a computer using an interpreter or the like.

Although one embodiment of the present invention as described above has been described by a limited embodiment and drawings, one embodiment of the present invention is not limited to the above-described embodiment, which is a general knowledge in the field of the present invention Those having a variety of modifications and variations are possible from these descriptions. Accordingly, one embodiment of the invention should be understood only by the claims set forth below, all equivalent or equivalent modifications will be within the scope of the invention idea.

FIG. 1 is a flowchart illustrating a representative image selection method when an image is duplicated according to an embodiment of the present invention.

2 is a diagram for describing duplication of a search target image according to an exemplary embodiment of the present invention.

FIG. 3 is a diagram illustrating an example of a process of extracting a plurality of identical images shown in FIG. 1.

4 is a flowchart illustrating a representative image selection method at the time of image de-duplication according to another embodiment of the present invention.

FIG. 5 is a diagram for describing a process of selecting a representative image using a term weight according to an embodiment of the present invention.

FIG. 6 is a block diagram illustrating a representative image selection system at the time of image de-duplication according to an embodiment of the present invention.

Claims (15)

Extracting a plurality of identical images; Exposing the plurality of identical images to a search object; Performing an image search on the input keyword based on the search target; And Exposing a representative image among the plurality of identical images when exposing a search result according to the image search; Representative image selection method during image deduplication, characterized in that it comprises a. The method of claim 1, Extracting the plurality of the same image, Comparing signatures of each image to determine image identity; Grouping a plurality of images determined to be identical to each other and extracting the plurality of identical images Representative image selection method during image deduplication, characterized in that it comprises a. The method of claim 2, The signature of each image is And a hash value of each of the images. The method of claim 1, Performing the image search, And when the input keyword is included in a word in a document including the image, the image is searched. The method of claim 1, Exposing the representative image, Selecting a representative image among a plurality of identical images in the search result; And Exposing the representative image by excluding an image excluding the representative image from the plurality of identical images from the search result Representative image selection method during image deduplication, characterized in that it comprises a. The method of claim 1, Calculating term weights for one or more words in each document including the plurality of identical images; And Scoring each of the plurality of identical images using the term weight More, Exposing the representative image of the plurality of the same image, And selecting the image having the highest score among the plurality of identical images as the representative image. The method of claim 6, Scoring the plurality of identical images, And assigning the determined term weight to the one or more words, and adding the term weights assigned to each word for each image to score the plurality of identical images. A computer-readable recording medium in which a program for executing the method of any one of claims 1 to 7 is recorded. A duplicate image extractor extracting a plurality of identical images; A search target manager to expose the plurality of identical images to a search target; A search performing unit which performs an image search on the input keyword based on the search target; And Search result providing unit for exposing a representative image of the plurality of the same image when the search results are exposed according to the image search Representative image selection system during image deduplication, characterized in that it comprises a. 10. The method of claim 9, The duplicate image extractor, A duplicate determination unit comparing signatures of each image to determine image identity; Redundant classification unit for grouping a plurality of images determined to be identical to each other Representative image selection system during image deduplication, characterized in that it comprises a. The method of claim 10, The signature of each image is And a hash value of each of the images. 10. The method of claim 9, The search performing unit, And when the input keyword is included in a word in a document including the image, the image is searched. 10. The method of claim 9, The search result providing unit, A representative image selecting unit selecting a representative image among a plurality of identical images in the search result; And An image exposing unit exposing the representative image by excluding an image except the representative image from the plurality of identical images from the search result Representative image selection system during image deduplication, characterized in that it comprises a. 10. The method of claim 9, A term weight calculator configured to calculate term weights for one or more words in each document including the plurality of identical images; And An image scoring unit that scores each of the plurality of identical images using the term weight. More, The search result providing unit, And the image having the highest score among the plurality of identical images is selected as the representative image. The method of claim 14, The image scoring unit, And assigning the at least one word to the at least one word, and adding the at least one weight to each word for each image to score the plurality of identical images.

Images