System and method for compiling images from a database and comparing the compiled images with known images

This patent application is a continuation of pending application Ser. No. 11/585,668 filed Oct. 24, 2006 by Rene Seeber, which is a continuation of U.S. Pat. No. 7,130,466, filed Dec. 21, 2000, and issued Oct. 31, 2006.

1. Field of the Invention

This invention relates generally to computer networks. More particularly, this invention provides for a system and method for searching and compiling from a database, such as the Worldwide Internet, images that have a specified visual content, and for determining if any of the compiled images are substantially similar to one or more known images.

2. Description of the Background Art

In its embryonic stage the Worldwide Internet provided a research-oriented environment where users and hosts were interested in a free and open exchange of information, and where users and hosts mutually trusted one another. However, the Internet has grown dramatically, currently interconnecting over 100,000 computer networks and several million users. Because of its size and openness, the Internet has become a target of trademark and service mark infringement or misuse. Virtually every trademark or service mark is available for unauthorized use on the Internet. Before connecting, companies balance the rewards of an Internet connection against risks of infringement of trademarks and servicemarks.

An entity\'s brands, trademarks, or servicemarks may be its most valuable asset. This is especially true with global intellectual property such as brands, trademarks, or servicemarks where integrity of the brand, trademark or servicemark is vital in new markets. Unfortunately, piracy of such intellectual property in many of these markets already costs leading corporations billions of dollars in lost sales annually, including new forms of piracy on the Worldwide Internet. Brand images (or look-alike marks) can be surreptitiously posted on web pages for selling fraudulent or unauthorized goods to a global market. If the presence of any brand, trademark or servicemark on the Internet becomes compromised, the result can be dilution of such any brand, trademark or servicemark, and ultimate loss of market share.

In the Worldwide Internet the number of web sites and the number of images increases daily by millions. Right now, there are expected to be more than 500 million images in the Internet. While searching for regular text in the Internet is known (e.g., commercial text search engines like Yahoo, Altavista, Lycos, etc.), searching solely for images is much more difficult. Presently, searching for images in the Internet is possible only by looking at an image name, e.g., “Clinton.gif,” or by looking at the text grouped around an image in a website (e.g., commercial sites like “richmedia.lycos.com,” Altavista image finder, etc.). It is believed that there is presently no feasible system to efficiently search for images in the Internet by specifying their visual content, because no computer system or computer method is presently available to detect the specified visual content of an image from all of the millions of images provided in the Internet.

Therefore, what is needed and what has been invented is a system and method for searching and compiling from a database, such as the Worldwide Internet, images that have a specified visual content, and for determining if any of the compiled images are substantially similar to one or more known images. What has been more specifically invented is a high-precision, automated visual detection service to protect global trademarks, servicemarks, and brands from infringement, dilution, or tarnishment by look-alike or imposter marks and brands on the Internet. The visual detection technology provided by the present invention finds a brand, trademark, or servicemark on Internet web pages, and also finds designs, symbols, shapes, and signs that closely resemble the brand, trademark or servicemark. The present invention also identifies logos within a larger picture and text within images.

The present invention broadly provides a system and method for searching and discovering from a database (e.g., the Worldwide Internet) an object (e.g. a logo, a trademark, etc.) which is confusingly similar with a known object. Broadly, an object crawler sweeps websites of the Internet by automatically following hyperlinks contained in the websites. On each website the object crawler identifies all objects and duplicates them by downloading them on servers of a temporary storage system. Broadly further, after the object are downloaded by the object crawler and stored on the servers of the temporary storage system, the visual content of the objects may be analyzed, such as by hundreds of parallel computers analyzing object content. This may be done in a massive parallel manner with hundreds of computers (e.g., three hundred computers or more). Each computer object operates an object analysis software component which processes one or more input objects and produces as output descriptive information in terms of text and numbers about what content is in the object(s). For each object the following information may be produced and stored: object size; “fingerprint” for efficient identification of substantial similar objects; all text contained in the object(s); “fingerprint” of each face contained in the object(s); information about the logos/trademarks contained in the object(s); and information about things and images contained in the object(s).

In one embodiment of the present invention, a graphical user interface is provided where the user may enter search criteria for the object to be searched. The search criteria to be entered in the graphical user interface may include one or more of the following search criteria: (i) one or more text strings that may be contained in the object including any image; (ii) one or more logo, trademarks or servicemarks selected from a list of predefined logos, trademarks or servicemarks that may be contained in the object including any image; (iii) one or more things or physical features or shapes selected from a list of predefined logos, trademarks or servicemarks that may be contained in the object including any image; (iv) one or more faces of facial templates that may be contained in the object including any image; and (v) one or more images that look substantially similar.

In another embodiment of the present invention, a system and method is provided for searching for an entity\'s logos, trademarks or servicemarks in objects and images in the Worldwide Internet. A known logo and/or trademark and/or servicemark is provided and is entered into the system of the present inventions; and the content of each object in the internet is compared with the known logo and/or trademark and/or servicemark to determine if there is any confusing similarity. If a confusingly similar logo and/or trademark and/or servicemark appears in the internet object, a reference to the internet object is stored as search results. After scrutinizing objects in the internet, the user may access the search results.

In yet another embodiment of the present invention, a system and method is provided for searching for faces of people or animals that are substantial identical to a known face. The system and method of embodiments of the present invention accept as input an object (e.g., a scanned photograph) that contains at least one face. Subsequently, the input face is compared with all faces in the internet objects (including images) using already computed face “fingerprints” available in storage. The result of the comparison is output in the form of a list of substantial identical objects (including images) that contain a face that is similar or substantially identical to the input face.

Embodiments of the present invention more specifically provide a method for discovering from a database (e.g., the Worldwide Internet) an object which is confusingly similar with a known object comprising: (a) searching (e.g. searching with a web crawler by following hyperlinks contained in web site elements) a database for objects; (b) providing a known object; and (c) determining if any object from the database is confusingly similar with the known object. The method preferably additionally comprises duplicating the objects from the database to produce duplicated objects; storing the duplicated objects to produce stored duplicated objects; and determining if any stored duplicated object is confusingly similar with the known object. The method further preferably additionally comprises determining the degree of similarity of any stored duplicated object with the known object. The objects may be selected from the group consisting of graphic images, videos, audio sounds and mixtures thereof. Each of the objects may be an intellectual property selected from the group consisting of logos, trademarks, service marks, and mixtures thereof. Determining if any object is confusingly similar with the known object further preferably comprises determining if all of the necessary metadata is available for any of the stored duplicated objects; and if not, the necessary metadata is developed for the stored duplicated objects. Determining if any object is confusingly similar with the known object further preferably comprises performing one or more of the following process steps: conducting an optical character recognition analysis on the object; conducting a facial analysis on the object; conducting a watermark analysis on the object; conducting a signature analysis on the object; and conducting an object similarity analysis on the object.

Embodiments of the present invention also more specifically provide a method comprising accessing a store that is storing duplicated objects from a database (e.g., an Internet database); and determining if any of the duplicated objects stored in the store are similar with a known object.

Embodiments of the present invention further also more specifically provide a computer-readable storage medium storing program code for causing a processing system to perform the steps of: searching a database for objects; duplicating the objects from the database to produce duplicated objects; storing (e.g., maintaining in memory or transferring into memory) the duplicated objects to produce stored duplicated objects; determining if any stored duplicated object is confusingly similar with a known object.

Embodiments of the present invention also provide for a system for discovering from a database an object which is confusingly similar with a known object comprising: a search engine for searching a database for objects; a duplicator coupled to the search engine for duplicating objects from the database to produce duplicated objects; a store coupled to the duplicator for storing the duplicated objects to produce stored duplicated objects; and determining means, coupled to the store, for determining if any stored duplicated object is confusingly similar with a known object. The system additionally preferably comprises determining the degree of similarity of any stored duplicated object with the known object.

The present invention further also provides a system for discovering from a database an object which is confusingly similar with a known object comprising: means for searching a database for objects; means for duplicating objects from the database to produce duplicated objects; means for storing the duplicated objects to produce stored duplicated objects; and means for determining if any stored duplicated object is confusingly similar with a known object. The system additionally preferably comprises means for determining the degree of similarity of any stored duplicated object with the known object.

The present inventions also provides a method for determining a degree of similarity between a known object and an object duplicated from a database comprising: duplicating an object from a database to produce a duplicated object; analyzing the content of the duplicated object (e.g., by assigning numbers for each pixel in the duplicated object) to produce a matrix of numbers; producing a model template from a known object; and comparing the model template of the known object with the matrix of numbers to determine the degree of similarity between the duplicated object and the known object. The method for determining a degree of similarity between a known object and an object duplicated from a database preferably additionally comprises one or more of the following process steps: providing a threshold degree of similarity to set a standard for confusingly similarity between the known object and the duplicated object; displaying the degree of similarity if the degree of similarity is at least equal to the threshold degree of similarity; and determining what region of the object the known object is located. The matrix of numbers is created in a RAM when the object (or image) is loaded from storage. The model template is computed and/or created automatically when the first search for object (e.g., a logo) is executed. The model template may be stored in a RAM. Each pixel consists of three numbers representing red, green, and blue. Color depends on algorithms. For example, in object or image searching, the colored image is converted into a grayscale image; subsequently, the actual analysis (or object/image detection) is performed on the grayscale image. The foregoing provisions along with various ancillary provisions and features which will become apparent to those skilled in the art as the following description proceeds, are attained by the practice of the present invention, a preferred embodiment thereof shown with reference to the accompanying drawings, by way of example only, wherein: