|
AUSTIN, Texas - A patent filed by Google was published Thursday by the World Intellectual Property Organization, an agency of the United Nations, holds the possibility of expanding the scope of  what can be searched online, by recognizing text within the images that make up videos such as those on YouTube. The patent, filed with the WIPO in June 2007 by Google and two of its computer scientists responsible for the company's Google Street View maps and Google Book Search, Adrian Ulges and Luc Vincent, envisions people being able to use a Google search to find what is on the shelves of the local grocery store or what is on display at the nearest museum, complete with related advertisements served up by Google. An Advanced Form of Optical Character Recognition The patent sought by Google is for a method of using computers to find any text within images, enhancing it and then performing Optical Character Recognition on any text found, essentially turning the text from a vacation video of a busy city street into a searchable database of business names, street signs, window signs or even the words on the tee-shirts of people passing by on the street. A person could use the technology sought in the patent to search for all the Greek statues held in the nearest museum, a type of search used as an example in the patent application. If the museum has been captured on video, Google's software method could extract the text from the signs describing each item and present the requested list of Greek statues to the searcher, including the location of each which could be presented on a custom museum map. The technology aims to identify "text of varying size, orientation, and typeface," according to the patent application. Google Text in Picture Patent Published In the patent application the technology is described in detail. "Extracted image text can be stored for use an in image search application. Image search applications are used to retrieve and present images for users, for  example, according to one or more search terms," the WIPO patent application states. "Each image is associated with keyword search terms, for example, derived from an image caption, image metadata, text within a predefined proximity of the image, or manual input. Additionally, image search application can include the text extracted from within the images to identify keywords associated with the image. Thus, the text within the image itself can be used as a search parameter," the application adds. Google's proposed technology for finding text within video includes a method for giving greater importance to certain text and less to others, which could be used to filter out text Google sees as misleading or inappropriate. "... a particular weighting is [...] applied to image text. For example, matches to image text can be given greater (or smaller) weight in the search results over text within a caption or otherwise associated with the image, which can be misleading. Alternatively, image text can be used to filter search results to eliminate particular images from a search result according to one or more predefined keywords (e.g., to reduce the retrieval of inappropriate images, spam filtering, etc.)," according to the patent application. Highlighted Text Possible The patent application includes methods helping make text within video stand out, such as highlighting or underlining text if needed, and for displaying relevant advertisements. "[...] in one implementation, the image is presented along with one or more advertisements. The advertisements can be selected based on the content of one or more search terms provided by the user," the application states. With the technology, Google may not even require a person to type in search terms, instead using either lists or images to find what a person is looking for within videos. Using the museum example Google notes, "The user can search for a particular object or use keywords to identify objects associated with an area of interest [...]. Alternatively, the user can browse [...] the museum to learn about the various objects." An example image from the Google's WIPO patent application is presented on the right, showing several instances of identifiable text noted by numbers. Contextual Advertisements Should the patent, which is being sought worldwide, be granted, Google may be able to display advertisements for a business that is seen and identified in a video clip on YouTube, or to show ads for what it calls "alternative businesses" in the surrounding area. In much the same way that Google's Street View maps have been generated using cameras mounted on vehicles driving in various U.S. cities, Google envisions cameras operated by "a small motorized vehicle or robot" indexing the contents of a business such as a grocery or hardware store, which could someday allow shoppers to find the lowest prices on groceries or to make sure a certain item is in stock before visiting. "The aisles of the store are traversed and images of products are captured," according to the patent application. Google has faced privacy related complaints for its Street View maps, and should the patent published Thursday be granted, it is likely to face further scrutiny. Google appears to see these methods as part of a plan to help make daily tasks such as shopping easier. "[...] A user searching for a particular location or business is provided with search results as well as additional results associated with the location or business," the application states. "Images of the destination location as well as the associated results are presented to the user. Other information retrieved from the particular images can optionally be presented to the user as well. For example, business hours extracted from the image can be shown," Google added. Implementation and Approval The technology has the potential to identify such things as tattoos of people walking in public, license plate numbers, or the lowest gas prices. "[...] digital images can illustrate landscapes, people, urban scenes, and other objects," according to the patent application. Not all of the patents sought by Google are granted, and of those that are not all are implemented. "[...] We file patent applications on a variety of ideas that our employees come up with. Some of those ideas later mature into real products or services; some don't," according to an unnamed Google spokesperson quoted in a recent Information Week article. "Prospective product announcements should not necessarily be inferred from our patent applications," the spokesperson added. Google's patent application, "Recognizing Text in Images," is in the initial publication release stage and faces additional study by the WIPO before being granted. Related Links:
|
5:47 pm on Jan. 4, 2008 (utc 0)