JPH11250106A - Method for automatically retrieving registered trademark through the use of video information of content substrate - Google Patents

Abstract

PROBLEM TO BE SOLVED: To retrieve impurity video and similar video by representing the geometric form of a trademark image by a set of Zernike moments or pseudo moments. SOLUTION: When a remote user inputs a trademark through an inquiry video inputting device 1 on a direct system or inquires through a network in the retrieving process of the trademark video progressed by on-line operation, the Zernike moments or the pseudo Zernike moments of the impurity video are extracted on-line by a Zernike moment feature vector extractor 3 to be stored as a feature vector of the inquiry video. Similarity between the inquiry video and the feature vector stored in a data base is calculated by a feature vector comparator 5 with respect to all the videos stored in the data base. Then, after aligning in the order of similarity, a specific number of videos provided for the remote user through a video output device 8.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method for automatically searching for a registered trademark using content-based video information, and more particularly to a method for using a registered trademark video itself on a digital video database in which registered trademarks are stored. The present invention relates to a method for automatically searching for a desired registered trademark.

[0002]

2. Description of the Related Art In order to secure the right to a trademark, the number of applications for registering a trademark is rapidly increasing. Therefore,
In the process of registering a new trademark or creating a trademark design, it is necessary to introduce an efficient method for confirming discrimination from existing trademarks. In other words, there is a growing need to quickly search and browse existing registered trademarks using the trademark image itself. At present, the demand for such content-based video search has been greatly increased due to a rapid increase in the use of digital video databases. However, it is very difficult to automatically analyze the contents of the video and search for a video similar to the video given as the question.

The current research trends and technologies are basic research on automatic retrieval of photographs, moving images, etc., and are in fact commercialized in a limited range. Research on automatic retrieval of registered trademarks is also in full-scale practical use. It is still in the early stages of system development.

[0004]

SUMMARY OF THE INVENTION According to the present invention, a user seeking to search whether a registered trademark similar to the trademark to be filed has already been filed has a trademark image search function on a network. Provides a method to configure a system that has the function of analyzing the transmitted trademark image, automatically searching for similar trademark images, and transmitting it to the user, by transmitting the transmitted trademark image to the installed server. It has its purpose to do.

[0005]

SUMMARY OF THE INVENTION In order to achieve the above-mentioned object, the present invention provides a method for automatically retrieving a trademark image according to the content or form of the trademark. Zernik
e) Pseudo-Zernik
e) To provide an automatic trademark search method including a step of expressing a set of moments.

Here, the Journey moment is a complex orthogonal moment whose absolute value is rotation invariant. References on jarniki moments include 1) A. Khotanzad and YHHong, [Invariant image
recognition by Zernike moments, IEEE Trans.on Pat
tern Analysis and Machine Intelligence, Vol.12, pp.
489-498, 1990] and 2) MRTeague, [Image analysis
via the general theory of moments, J. Opt. Soc, Am.,
1980] and 3) A. Khotanzad and YHHong, [Rotntio i
vnrint imnge recogitio using fentures selected vin
n systemnticmethod, Pntter Recognitio, Vol.23, p
p.1089-1101, 1990] and 4) CHTeh and RTChi, [O
imnge n nlysis by the methods of momets, IEEE Trn
so Pntter A nlysis nd Mnchie I tellige ce, Vol.1
0,1988.].

In the above document 4), CHTeh is:
Various moments are compared from the viewpoints of sensitivity to image noise, information redundancy, and ability to express images. This study confirms that jarniki and pseudo-jarniki moments are superior to general moments, lege dre, rotational moments, and complex moments.

In the present invention, the polymerization of the magnitude of the jarnik moment or the quasi-jarnik moment (Weig)
hted Sum) may be used to measure the similarity between the two trademark images. Specifically, using the statistical properties (eg, mean or standard deviation) of the magnitude of the jarnik moment or the quasi-jarnik moment, the jarnik or quasi-jarnik moment of each of the two trademark images is normalized. , Then the similarity distance between the two
tance).

The step of measuring the degree of similarity includes the steps of inputting a query trademark image, extracting a jarnik moment or a pseudo jarnik moment of the query trademark image using a shape vector, and The method may include calculating a similarity by comparing a vector with a shape vector of the query mark image, and outputting a registered trademark image aligned with the calculated similarity.

[0010]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, the present invention will be described in detail with reference to the attached drawings.

FIG. 1 is a block diagram of an automatic search system for registered trademarks using content-based video information according to the present invention. FIGS. 2 and 3 are registered trademarks using content-based video information according to the present invention. FIG. 4 is a search result diagram obtained by an embodiment of the automatic search system of FIG. 1, and FIG. 4 is a flowchart of an automatic search system of a registered trademark using content-based video information according to the present invention.

A general video information retrieval system extracts a feature vector from a video given as a query, extracts a feature vector from videos stored in a database, and determines the similarity between them. An image similar to the question image is searched for by comparing.

In this case, a method of extracting a feature vector as a criterion of similarity is very important, but in the present invention, a Zernike moment extracted from registered trademark video data as a feature vector representing similarity between registered trademarks is used. Or use a pseudo-Zernike moment.

The similarity is determined based on the question vector and the feature vector of the video belonging to the video database of the registered trademark, and the registered trademark is arranged and provided to the user based on the similarity. It also creates a file for automatically searching for a registered trademark on a general network including the Internet, transmits search trademark data, and transmits search results.

The system for automatically searching for a registered trademark using content-based video information according to the present invention, as shown in FIG. 1, uses a database construction process that is performed offline and a trademark video that is processed online. In the search process.

Here, the Jernicky moment feature vector extractors 3 and 4 and the feature vector comparator 5 provided in the server are constructed as processes by the information processing apparatus executing a predetermined program. Also,
The trademark video input device 2 is realized by a mouse or a digitizer included in the information processing device. The feature vector database 6 and the trademark video database 7 are realized by a storage device such as a hard disk included in the information processing device.

The trademark question input device 1 and the retrieved trademark video output device 8 provided in the client are provided with a mouse, a digitizer, and the like included in the information processing device when the information processing device executes a predetermined program. , Realized by the display.

Further, the information processing device functioning as a client and the information processing device functioning as a server are:
It is assumed that they are connected to each other via a network such as an AN and the Internet.

Referring to FIG. 1, in the process of constructing a database, which is performed in an off-line operation, when a trademark image is input to the database by the trademark server image input unit 2, a Zernike moment or a pseudo-Janiki is generated. -Zernike) Moments are extracted as feature vectors (4) and stored in the database along with the video (6,7).

In the process of retrieving a trademark image performed online, if a remote user directly inputs a trademark image through a question image input device on the system or asks a question using a video through a network (1), The Journey or pseudo Journey moment is extracted online (3) and stored as a feature vector of the question video.

The similarity between the question video and the feature vectors stored in the database is calculated for all the videos stored in the database (5), and after sorting them in the order of similarity, a specific number of The image is provided to the remote user through the image output device (8).

A video input device (1) for a remote user has a function of drawing a picture using a mouse or a digitizer, a function of selecting one of prototype trademarks provided by a server, and a function of a user. The function of selecting a picture stored in the hard disk and reading it with the video input device is supported.

The remote user can read an existing image and modify it, or perform a query by directly drawing a new picture. The image provided by the server to the remote user as a result of the query is used as a prototype trademark provided by the server, and can be used as it is or transformed and re-queryed.

The video output device (8) for the remote user provides the user with a typical trademark pattern as a prototype trademark at the time of the user's initial connection, and the user selects one of these patterns. If selected, it has the function of passing it to the user's video input device (1).

When the server responds to the user's query with the result, the video output device shows the result to the user and allows the user to also use the result as a prototype trademark. Be able to ask more detailed questions.

The Journey or pseudo Journey moment feature vector extractors (3 and 4) extract the Journey moment or pseudo Journey moment from the given image as follows.

The Zarniki radiation function vector is defined as follows:

[0028]

(Equation 1)

On the other hand, in the case of pseudo-Journiki, the radiation function vector is defined as follows.

[0030]

(Equation 2)

The Journey moment extracted from the image I (ρ, θ) expressed in polar coordinates can be expressed as follows.

[0032]

(Equation 3)

Here, V (ρ, θ) is a Journey basis function vector obtained by the following equation.

[0034]

(Equation 4)

The Journey moment A extracted from an image is a complex number, and when its absolute value is taken, a feature vector z used in the present invention is extracted as in the following equation.

[0036]

(Equation 5)

This is a vector consisting of z _nm as follows:

[0038]

(Equation 6)

An arbitrary number of feature vectors can be extracted.
Extraction of various features increases accuracy, but requires a lot of calculation time and memory. Therefore, the number of feature vectors is determined in consideration of accuracy, calculation time, and storage capacity.

The feature vector extracted in this way is a feature vector that well reflects the pattern of the pattern in the video, is resistant to noise or deformation, and well reflects the form of the geometric pattern. It is suitable for use in feature vectors for measuring degrees.

The feature vector comparator compares the Journey moment (3) extracted from the image input by the trademark input device (1) of the remote user of FIG. 1 with the feature vector stored in the database. To determine the similarity and provide the result to the trademark image output device (8) of the remote user.

When the user makes a re-question using the search result, the currently calculated similarity and the previous similarity are added, and the previous question and the current question are combined and used. To be able to

As described above, according to the present invention, in terms of sensitivity to image noise, redundancy of information, and ability of image expression, the jarnik and the pseudo jar exhibit superior levels as compared with general moments, rotational moments, and complex moments. The Niki moment is used. As a result, the image pattern can be effectively reflected on the feature vector, and a query image and a similar image can be searched.

FIGS. 2 and 3 show search results obtained by applying the present invention to a database composed of 3,000 trademark videos.

The trademark displayed on the leftmost side of FIG. 2 is a question mark, and the trademarks displayed on the right side are selected as trademarks similar to the question and selected from among the 30th or less rank among the trademarks output. It is indicated by.

Although the similarity judgment criterion is probably subjective, as a result of questioning 10 people engaged in this field and asking about 10 trademarks, the similarity ranking which is about 1% of the whole database is obtained. Of the trademarks searched within the 30th place, about 65% of the trademarks were actually similar to the question mark. That is, the accuracy of the search was about 65%.

The trademark displayed on the leftmost side of FIG. 3 is an original trademark, and the trademark displayed on the right side is a modification of the trademark. As a result of conducting an experiment as to whether the modified trademark can be searched from the former trademark by a question and answer assuming that the trademark is similar to the original trademark, the ratio of searching within the 30th rank of similarity is determined. Was about 92.5%.

In the search processing shown in FIGS. 2 and 3, the data of the trademark actually registered with the Korean Patent Office (600,000 trademark data provided for research at the Korean Patent Office) is used. It is applied to a database (a database composed of 3,000 trademark videos) constructed using graphic data provided by a company that designs a trademark. Here, the data of the trademark actually registered with the Korean Patent Office is used only for data that does not actually exist as a trademark, and is used for the purpose of the present application (on a digital video database in which the registered trademark is stored). To be able to automatically search for the desired registered trademark by using the registered trademark image itself for question and answer) (it is considered impractical to explain the utility of the present invention). This is because the. There is no other reason to use the data of the trademark actually registered with the Korean Patent Office, except to show the actual effect of the similarity judgment of the present invention.

FIG. 4 is a flow chart of an automatic registered trademark search system using content-based video information according to the present invention. The registered trademark automatic search system according to the present invention receives trademark image data from a user in a user mode.
(41). In the user edit mode, the user edits and inputs a desired trademark on a graphic editor.
(44). If the user mode is the input mode, a trademark image file that has already been edited is input and the file format is converted (42, 43). The image file input in the edit mode or the input mode is converted into a data form supported by an automatic search system of the registered trademark (45), and a jarnik moment or a quasi-jarnik moment form vector of the trademark image is extracted (46). ). The similarity is calculated by comparing the feature vector of the trademark image with the feature vector of the registered trademark image stored in the feature vector database (6) (47). The registered trademark images stored in the trademark image database (7) are arranged and output in the order of the calculated similarity (48, 49).

The present invention is not limited to the above embodiment, and can be implemented in various forms without departing from the gist of the present invention.

[0051]

As described in detail, according to the present invention, a pattern of an image can be effectively reflected on a feature vector.
A question video and a similar video can be searched. Such search results have a remarkable effect when compared with current content-based video search technology.

[Brief description of the drawings]

FIG. 1 is a block diagram of a registered trademark dynamic search system using content-based video information according to the present invention.

FIG. 2 is a view showing search results obtained by an embodiment of an automatic search system for registered trademarks using content-based video information according to the present invention;

FIG. 3 is a view showing a search result obtained by an embodiment of an automatic registered trademark search system using content-based video information according to the present invention;

FIG. 4 is a flowchart illustrating a registered trademark automatic search system using content-based video information according to the present invention;

[Explanation of symbols]

 1: trademark video query input device for remote users 2: video input device for server 3, 4: Journey moment feature vector extractor 5: feature vector comparator 6: feature vector database 7: video database 8: remote use Display of Trademarks Searched for Consumers

 ──────────────────────────────────────────────────続 き Continuing on the front page (72) Inventor Kim Mei-sung 161 Kejeong-dong, Yuseong-gu, Daejeon, Republic of Korea Inside the Korea Electronic Communications Research Institute (72) Inventor Jung-Heon Kejeong-dong, Yuseong-gu, Daejeon 161 Inside the Korea Institute of Electronics and Communications (72) Inventor Kim Eun-Ri South Korea Republic of Korea Ota Metropolitan City Yuseong-gu Kejeong-dong 161 Inside the Korea Electronics and Communications Research Institute (72) Inventor Kim Ryusei Korea Republic of Korea Yuseong-gu Ke 161 Yeongdong-dong

Claims (3)

[Claims] 1. A method for automatically retrieving a trademark image according to the content or form of the trademark, wherein the geometric form of the trademark image is represented by a set of Zernike moments or pseudo-Zernike moments. An automatic trademark search method comprising the step of expressing. 2. The method according to claim 1, further comprising measuring a similarity between the two trademark images using a weighted sum of the magnitude of the journic moment or the quasi-Journal moment. Automatic trademark search method. 3. The step of measuring the degree of similarity includes the steps of: inputting a query trademark image; extracting a jarnik moment or a pseudo jarnik moment of the query trademark image by a shape vector; 3. The method of claim 2, further comprising: calculating a similarity by comparing a vector with a morphological vector of the query mark image; and outputting the registered trademark image aligned according to the calculated similarity. Automatic trademark search method.