JPH1153541A - Method and device for discriminating image similarity and storage medium - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a similar image discriminating device for exactly discrimi nating the similarity between an image included in circulating contents and an original document image. SOLUTION: A similar image discriminating device (similar image discriminating server) 10 acquires contents from plural information providing servers 30 owning contents through a public network L. Then, the feature variables of the original document image and the contents are extracted by an image analysis part 13 and out of the contents, an object image to be compared with the original document image is specified. Afterwards, a similarity processing part 18 discriminates the similarity between the feature amounts of the object image and the original document image. When the similarity is low, an image converting part 19 approximates the feature amount by performing prescribed image converting processing to the original document image and discriminates the similarity again. When the images are similar, that object image is preserved together with its aquire destination information.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a technique for judging the similarity between feature values between images, and more particularly, to a content (multimedia that is presumed to contain an image identical or similar to an original work image). The present invention relates to an image processing technique applied to a system for determining the similarity to an original copyrighted work image by specifying the image from an electronic work made of information, the same applies hereinafter.

[0002]

2. Description of the Related Art Recently, with the development of network technologies such as the Internet, it has become easier to distribute digitized information in various forms. In such a situation,
The distribution of unauthorized images that ignore the rights of authors of digitized information also tends to increase, which is a bottleneck for building a sound advanced information society. For example, the creator of an image usually spends a great deal of time and effort creating the image as a form of expression of a certain idea or for a certain purpose. However, it has become extremely easy for an unauthorized user to create a secondary work using all or part of the image without the permission of the image creator. It is important from the viewpoint of protecting the copyright of the image creator to prevent such secondary works from being illegally distributed on a public network or the like. I have. For example, as a means for searching for a secondary work, an image similarity determination device (a system when performing distributed processing) has been conventionally proposed.

In this type of image similarity determination apparatus, a target image to be watched by an operator is extracted from a plurality of target images existing in an image database, a web on the Internet, or the like. When extracting the target image, in addition to visual observation, automatic search processing based on keywords added to each original work image in advance, sample the original work image itself and exemplify it, and automatically or manually search for similar images Has also been done. Whether the extracted target image is similar to the original copyrighted image is determined based on the distance between feature vectors expressing the shape and texture features of both images, or the positional relationship between image components is determined. Judgment is performed by matching represented by a graph structure or the like.

[0004]

As described above, in the conventional image similarity determination apparatus, a keyword or a sample image is used as a search key, and a target image closest to the keyword is selectively extracted from an image database or the like. I have. By the way, when the distribution monitoring of content is performed using the image similarity determination device, the secondary work usually corresponds to the target image, and the original work image corresponds to the search key, but if the image is a digital image, Since the gradation conversion, filtering, rotation, irreversible compression, cutout, and the like of the image are extremely easy, it is difficult to determine the similarity between the two images when the image conversion processing is performed on the target image. However, in the conventional image similarity determination device, since the search process is performed on the assumption that the target image has not been subjected to the image conversion process, the content includes an image that is the same as or similar to the original copyrighted image. It was difficult to determine whether or not it was accurate.

It is an object of the present invention to provide a computer-based image similarity determination method capable of accurately determining whether or not an image identical or similar to an original work image exists in distributed content. Another object of the present invention is to provide an improved image similarity determination device capable of accurately monitoring content distribution by increasing the accuracy of similarity determination between images, and a general-purpose image similarity determination device. An object of the present invention is to provide a recording medium to be realized by a computer.

[0006]

An image similarity determination method according to the present invention for solving the above-mentioned problems is a method for determining the similarity between a plurality of digitized images, which is the same as or similar to the original work image. A process of specifying a target image to be compared with the original work image from the content estimated to include the image, and at least one of the specified target image and the original work image, enlargement, reduction, rotation, A process of performing image conversion processing such as color tone conversion to approximate the feature amounts of both images, and a process of comparing the feature amounts of both images after the image conversion process to generate information representing the similarity between the images. It is characterized by including.

Further, the image similarity determination apparatus of the present invention for solving the above-mentioned other problems is presumed to include a digitized original work image and an image identical or similar to the original work image. Image acquisition means for acquiring content and
A target image specifying unit that specifies a target image to be compared with the original work image in the acquired content, and compares the specified target image and the feature amount of the original work image to determine the similarity between the images. A similarity determination unit configured to generate the information to be represented, and an output unit configured to output the information representing the similarity together with the acquisition source information of the target image.

In a more preferred embodiment, at least one of the specified target image and the original copyrighted image is subjected to image conversion processing such as enlargement, reduction, rotation, and color tone conversion to more closely approximate the feature values of both images. The image similarity determination device is further provided with an image editing means for causing the image similarity to be determined. In this case, the similarity determination unit is configured to compare the feature amounts of both images after the image conversion processing.

The similarity determination means determines the similarity between images through, for example, the following comparison processing. (1) An orthogonal transformation based on a wavelet transform coefficient is performed on each of the target image and the original copyrighted work image, and the characteristic amounts of the image waveforms obtained by analyzing the distribution characteristics of the wavelet transform coefficients are compared. (2) A predetermined area division is performed on each of the target image and the original copyrighted work image, and characteristic amounts of image shapes obtained by analyzing characteristic information including an image contour, a texture, and an area position of each divided area are compared. I do.

Another image similarity determination device of the present invention is a device connected via a communication path to a content holding device holding content including an image, and is the same as a predetermined digitized original work image. Or image acquisition means for acquiring a content presumed to include a similar image from the content holding device and acquiring the original copyrighted image, and a target image to be compared with the original copyrighted image in the acquired content. Target image specifying means for specifying the same; similarity determining means for comparing the specified target image and the feature amount of the original copyrighted work image to generate information indicating similarity between the images; and information indicating the similarity. And output means for outputting the target image together with the acquisition destination information of the target image, and monitors the image in the content distributed through the content holding device. It is characterized in.

In this case, the similarity determination means determines a target image whose difference in the feature amount is equal to or less than a predetermined threshold value as a similar image, and identifies identification information of a content including the target image together with its holding destination information. Configure to accumulate. The image processing apparatus further includes a presentation unit that visualizes and presents the content including the target image determined to be the similar image, so that the operator can easily confirm the presence or absence of the similar image.

A recording medium according to the present invention for solving the other problems is a computer-readable recording medium in which a program for causing a computer to execute the following processing is recorded. (1) a process of acquiring a digitized original work image and a content presumed to include the same or similar image as the original work image, (2) a comparison of the acquired content with the original work image (3) At least one of the specified target image and the original copyrighted image is subjected to image conversion processing such as enlargement, reduction, rotation, and color tone conversion, and (4) a process of comparing the feature values of both images after the image conversion process to generate information indicating a similarity between the images, and (5) a process of generating information indicating the similarity between the images. The process of outputting the target image together with the acquisition source information.

[0013]

Embodiments of the present invention will be described below in detail with reference to the drawings. (First Embodiment) FIG. 1 is a functional block diagram showing an embodiment in which the present invention is applied to a similar image search system. The similar image search system 1 includes a similar image determination server 10 that can access a plurality of information providing servers 30 functioning as content holding devices.

The similar image determination server 10 includes a communication control unit 11, a data input unit 12, an image analysis unit 13, an image waveform feature calculation unit 14, and a communication control unit 11, which are formed by the server body reading and executing a predetermined program. Image shape feature calculation unit 15, target image specification unit 16, feature vector calculation unit 1
7, a similarity processing unit 18, an image conversion unit 19, and a determination result output unit 20. The program is usually stored in an internal storage device or an external storage device of the server main body, and is read and executed as needed. However, a recording medium separable from the server main body, for example, a CD-ROM. It may be stored in a ROM or FD, read at the time of use, installed in the internal storage device or the external storage device, and provided for execution at any time.

The communication control unit 11 is a publicly known one that controls communication with the public network L. In this embodiment, control is performed such that a desired content is received through the public network L and input to the data input unit 12.

The data input unit 12 fetches an original work image from, for example, a scanner, a web, or an image database (not shown), and inputs the image to the image analysis unit 13 together with the content input from the communication control unit 11.

The image analysis unit 13 includes an image waveform feature calculation unit 1
4 and an image shape feature calculation unit 15, which analyzes the input original work image and content and extracts data representing the feature amount of each image. Image waveform feature calculation unit 1
Numeral 4 is for analyzing a distribution feature using the color tone information of each image as a wavelet transform coefficient and calculating a waveform feature amount. Further, the image shape feature calculation unit 15 divides each image into regions, analyzes the contour features, texture features, and region positions of each divided region, and calculates the shape feature amount of the image. The analysis result by the image analysis unit 13 becomes information such as the center of gravity relating to the degree of concentration of pixels in each image, the degree of concentration of contour components, changes in luminance, and peak values of wavelet transform coefficients.

The target image specifying unit 16 specifies an image (target image) to be subjected to the similarity determination from the content based on the analysis result from the image analyzing unit 13. The feature vector calculation unit 17 calculates the feature vectors of the original copyrighted work image and the target image specified by the target image specifying unit 16, respectively. At this time, the analysis result of the image analysis unit 13 is used. The similarity processing unit 18 performs a matching process between the feature vectors of the original copyrighted image and the target image, and determines the similarity between the images based on the inner product of the distance between the two. In addition,
The use of the distance between feature vectors as a measure of similarity determination between images is merely an example, and the similarity between images may be determined by image pattern matching processing or so-called “image watermarking”. . The image conversion unit 19 performs image conversion processing such as enlargement, reduction, rotation, and color tone on at least one of the original copyrighted image and the target image as needed. At this time, it is desirable to display the result of the image conversion process on a display device (not shown) so that the operator can confirm the result. The determination result output unit 20 outputs information on the similarity between the original copyrighted image and the target image based on the determination result by the similarity processing unit 18.

Next, the processing procedure of the similar image determination server 10 for detecting an image that is the same as or similar to the original copyrighted work image from a plurality of distributed contents will be specifically described with reference to FIG.

The similar image determination server 10 obtains an original work image from the data input unit 12, accesses the plurality of information providing servers 30 through the communication control unit 11 using the original work image as a search key, and Is acquired (step S101). here,
It is assumed that one content has been acquired. After that, the original work image and the content are input to the image analysis unit 13, and the feature amount of each image is analyzed (step S102). Specifically, a waveform feature value and a shape feature value are calculated. Then, the analysis result (calculation result) is input to the target image specifying unit 16, and the target image in the content is specified (step S1).
03).

When the target image is specified, the image analysis unit 13
In step S104, the image characteristics are analyzed again, and the feature vector of each image is calculated by the feature vector calculation unit 17 from the obtained feature amounts of the target image and the original work image (step S104). The calculated feature vector is input to the similarity processing unit 18.

The similarity processing unit 18 performs a matching process between the target image and the original copyrighted image based on the input feature vector (step S105). The matching process in this case, specifically, calculates the distance between the feature vectors, regards the function of the reciprocal of the calculated value or the inner product value as similarity, and if the similarity is equal to or greater than a preset threshold,
It is determined that the two images are similar. When it is determined that the two images are similar (Step S106: Yes), the similarity processing unit 1
8 passes the determination result on the target image to the determination result output unit 20 (step S107). Judgment result output unit 20
Displays the content including the target image and the original work image on a display device (not shown). This allows the operator to recognize that the acquired content includes the original work image, that is, the content is a secondary work of the original work image. When the content is acquired through its name and acquisition destination information, for example, via the Internet, the content name and the URL (Uniform Resource Locator) name are stored in a specific file. As a result, it is possible to specify a pursuit destination when the content is illegally used.

If it is determined that the two images are not similar as a result of the matching processing in step S105 (step S106: No), the similarity processing unit 18 sets an image switching flag to at least one of the two images, for example, the original copyrighted work image. The original work image is set and passed to the image conversion unit 19. The image switching flag is initially off, and is updated to be on after the image conversion processing unit 19 performs image conversion processing in a direction that approximates the feature amount of both images (step S108: off, S10
9, S110). After the image conversion processing, step S104
Execute the subsequent processing again.

On the other hand, if the image switching flag is ON in step S108, that is, if the original image is not similar to the target image even after the image conversion processing is performed (step S108: ON), the image is currently specified. The process returns to step S102 in order to specify another target image in the content again in place of the current target image. This process is repeated by the number of target images. When it is determined that all the target images are not similar to the original copyrighted work image, the acquisition of another content and the processing after step S102 are executed again.

In the above example, the image conversion processing is performed on the original copyrighted image, but the same image conversion processing is performed on the target image instead of the original copyrighted image or together with the original copyrighted image. In this way, the feature amounts of both images may be approximated.

As described above, in the similar image search system 1 of the present embodiment, the image conversion unit 19 performs gradation conversion, filtering, rotation, irreversible compression, and clipping on at least one of the original work image and the target image. And so on, so that the feature amount comparison between the images is performed. Therefore, even if the content is a secondary work obtained by converting the original work image, the original work image It becomes easy to detect the correlation with This makes it possible to quantitatively determine the unauthorized use of the content.

(Second Embodiment) The similarity determination device of the present invention can be implemented as a so-called stand-alone device. In this case, the similar image determination device includes a data input unit 12, an image analysis unit 13, and the like, similar to the similar image search system 1, to a stand-alone computer device having a display device for directly presenting information to an operator.
An image waveform feature calculation unit 14, an image shape feature calculation unit 15, an image identification unit 16, a feature vector calculation unit 17, a similarity processing unit 18, an image conversion unit 19, and a determination result output unit 20 are provided. .

The similar image determination apparatus differs from the similar image determination server 10 in that the similar image determination apparatus does not include a functional block for performing communication control. In such a case, the image input from the operator can be directly received, so that an alternative is possible. Therefore, the same effect as in the first embodiment can be obtained.

[0029]

As is apparent from the above description, according to the present invention, the correlation between the original work image and the original work image is detected from the contents presumed to include the same or similar image as the original work image. Since the accuracy of the similarity determination of the content can be improved, it is possible to quantitatively determine whether or not the content is illegally used. In addition, since the distribution monitoring of the contents can be accurately performed, the proper and smooth distribution of the contents is promoted, and the operation management of a sound network environment becomes possible.

[Brief description of the drawings]

FIG. 1 is a functional block configuration diagram of a similar image search system to which the present invention is applied.

FIG. 2 is an explanatory diagram illustrating a procedure of an image similarity determination process in the similar image determination server according to the embodiment.

[Explanation of symbols]

 Reference Signs List 1 Similar image search system 10 Similar image determination server 11 Communication control unit 12 Data input unit 13 Image analysis unit 14 Image waveform feature calculation unit 15 Image shape feature calculation unit 16 Image specification unit 17 Feature vector calculation unit 18 Similarity processing unit 19 Image Conversion unit 20 Judgment result output unit 30 Information providing server

Continuation of the front page (72) Inventor Taichi Nakamura 3-3-3 Toyosu, Koto-ku, Tokyo NTT Data Communications Corporation

Claims (9)

[Claims] 1. A method for judging the degree of similarity between a plurality of digitized images, comprising comparing an original work image from a content estimated to include an image identical or similar to a predetermined original work image. A process of specifying a target image to be processed; and performing image conversion processing such as enlargement, reduction, rotation, and color tone conversion on at least one of the specified target image and the original copyrighted image to approximate feature amounts of both images. An image similarity determination method using a computer, the method including a step of comparing the feature amounts of both images after the image conversion processing to generate information representing a degree of similarity between the images. 2. An image acquisition means for acquiring a digitized original work image and a content presumed to include the same or similar image as the original work image, and A target image specifying unit for specifying a target image to be compared with the crop image; and a similarity for generating information representing the similarity between the images by comparing the feature amount of the specified target image with the feature amount of the original work image. An image similarity determination device, comprising: a determination unit; and an output unit configured to output information indicating the similarity together with information on a source of the target image. 3. Enlargement, reduction, rotation, and / or rotation of at least one of the specified target image and the original work image.
The image processing apparatus further includes an image editing unit that performs image conversion processing such as color tone conversion to more closely approximate the feature amounts of the two images, and the similarity determination unit compares the feature amounts of the two images after the image conversion process. 3. The image similarity determination device according to claim 2, wherein: 4. The image waveform obtained by performing the orthogonal transformation based on a wavelet transform coefficient on each of the target image and the original copyrighted work image, and analyzing a distribution characteristic of the wavelet transform coefficient. The image similarity determination device according to claim 2 or 3, wherein the feature amounts are compared. 5. The similarity determination means performs a predetermined area division on each of the target image and the original work image,
4. The image similarity determining apparatus according to claim 2, wherein the feature amounts of the image shapes obtained by analyzing the feature information including the image contour, the texture, and the region position of each divided region are compared. 6. A device connected via a communication path to a content holding device holding content including an image, and is presumed to include an image which is the same as or similar to a digitized original original work image. Image acquisition means for acquiring the original copyrighted work image from the content holding device, and target image identification means for identifying a target image to be compared with the original copyrighted work image in the acquired contents, A similarity determination unit configured to compare the feature amounts of the specified target image and the original copyrighted image to generate information indicating the similarity between the images; and output the information indicating the similarity together with the acquisition source information of the target image. Output means for performing image monitoring in the content distributed through the content holding device. Setting device. 7. The similarity determination means determines a target image in which the difference in the feature amount is equal to or less than a predetermined threshold value as a similar image, and stores identification information of content including the target image together with its holding destination information. The image similarity determination apparatus according to claim 6, wherein: 8. The image similarity determination apparatus according to claim 7, further comprising a presentation unit that visualizes and presents a content including the target image determined as the similar image. 9. A process of acquiring an electronic original work image and a content presumed to include an image identical to or similar to the original work image, and comparing the acquired content with the original work image A process for specifying a target image to be processed, and performing image conversion processing such as enlargement, reduction, rotation, and color tone conversion on at least one of the specified target image and the original copyrighted image so as to improve the feature amount of both images. A process of approximating, a process of comparing the feature amounts of both images after the image conversion process to generate information indicating a similarity between the images, and outputting the information indicating the similarity together with the acquisition source information of the target image And a computer-readable recording medium on which a program for causing a computer to execute the process is recorded.