Image analysis – Pattern recognition – Feature extraction
Reexamination Certificate
1998-08-14
2001-12-04
Johns, Andrew W. (Department: 2721)
Image analysis
Pattern recognition
Feature extraction
C382S203000
Reexamination Certificate
active
06327388
ABSTRACT:
BACKGROUND AND SUMMARY OF THE INVENTION
The present invention relates generally to image analysis and electronic information identification. More particularly, the invention relates to a computerized system for locating and identifying logos in document images.
In this information age, we are constantly bombarded with electronic files and information. Some call this the information explosion. Much of this information derives from image-based systems that generate bitmapped images. To help cope with the information explosion we need many tools, including ones to distinguish electronic files and documents based on information contained in bitmapped images.
Image-based systems present a major challenge to information retrieval because the data is difficult to search. Whereas information retrieval technology is fairly well advanced in coded character-based systems, these retrieval techniques do not work in image-based systems. That is because image-based systems store information as bitmap data that correspond to the appearance of the printed page and not the information content of that page. Traditional techniques require the conversion of bitmap data into text data, through optical character recognition (OCR) software, before information retrieval systems can go to work. Unfortunately, optical character recognition software is computationally expensive, and the recognition process is rather slow.
In addition, often important information may be encoded graphically, in a form that cannot be readily converted into text. Corporate logos represent an important example of such information. While some corporate logos do include letters that can be converted using OCR, many do not. It would therefore be quite helpful to be able to identify logos, regardless of whether they contain recognizable letters, to allow the computer to search and retrieve information based on the presence of the logo. This would allow one to organize documents and files by logo, if desired. Organizing information on this basis can be quite helpful, even if character-based organization is also possible.
For example, suppose a user wanted to locate any document received from ABC Corporation. Normally this would be performed as a text search for the letters “ABC”, and this text search would yield any three letters “ABC” that are connected together as a single entity. However, if ABC Corporation uses a highly stylized logo instead of easily recognizable letters, the text search technique may be of little value. Conversely, such a text search would also uncover other instances of the letters “ABC,” possibly having nothing to do with a document received from ABC corporation. A text search would also retrieve newspaper articles about ABC corporation, for example. In this example a logo search would be highly desirable. It could locate the document of interest without the foregoing problems.
The present invention recognizes that there will be vast amounts of data that are in bitmap or image format, and that users will want to search this information, just as they now search text-based systems. Instead of converting the entire document from image format to text format, the present invention analyzes the bitmap data in its native format, to extract regions within the image data that correspond to the most likely candidates for logos. The invention recognizes that the logos may be in any position in the document and at different skew angles. The invention scans the document, seeking a bitmap that is similar to the bitmap of the logo in question.
In accordance with the invention, a document image is matched with a database of logo images stored in a computer readable memory as a logo bit pattern and as predefined sets of parameters. The predefined sets of parameters are indicative of shape and mean density. Connected component analysis is performed on the bit pattern of a document image to identify a plurality of regions of connected components in the document image and to generate region parameter data that include parameters indicative of shape and mean density.
The parameters of a document region are compared with the parameters of a database of logo images to identify a plurality of logo match candidates. The document regions are then scaled and aligned to correspond to the size and skew of the logo match candidates. The result of the document regions being scaled and aligned is to create a number of adjusted regions that also have corresponding adjusted bit patterns. A bitwise comparison is performed between the adjusted bit patterns with their corresponding adjusted regions and the bit patterns of the logo match candidates to select a best match and store the best match selections in a ranked logo list database.
For a more complete understanding of the invention, reference may be had to the following specification and to the accompanying drawings.
REFERENCES:
patent: 4186378 (1980-01-01), Moulton
patent: 4461028 (1984-07-01), Okubo
patent: 4817166 (1989-03-01), Gonzalez et al.
patent: 4972499 (1990-11-01), Kurosawa
patent: 5287275 (1994-02-01), Kimura
patent: 5351314 (1994-09-01), Vaezi
patent: 5414781 (1995-05-01), Spitz et al.
patent: 5436983 (1995-07-01), Bernzott et al.
patent: 5438628 (1995-08-01), Spitz et al.
patent: 5452094 (1995-09-01), Ebner et al.
patent: 5479537 (1995-12-01), Hamashima et al.
patent: 5555362 (1996-09-01), Yamashita et al.
patent: 5588027 (1996-12-01), Wang
patent: 5680479 (1997-10-01), Wang et al.
patent: 5699453 (1997-12-01), Ozaki
patent: 5703962 (1997-12-01), Niki et al.
patent: 5748865 (1998-05-01), Yamamoto et al.
patent: 5751849 (1998-05-01), Ikeda
patent: 5757957 (1998-05-01), Tachikawa
patent: 5767978 (1998-06-01), Revankar et al.
patent: 5774579 (1998-06-01), Wang et al.
A. Lawrence Spitz, “Logotype Detection in Compressed Images using Alignment Signatures”, Daimler Benz Research and Technology Center, pp. 303-310.
M.Y. Jaisimha, “Wavelet Features for Similarity based Retrieval of Logo Images”, Apr. 1996, vol. 2660, pp. 89-100.
David Doermann et al., “Logo Recognition”, Center for Automation Research, University of Maryland, Oct. 1993, pp. 1-28.
Lopresti Daniel P.
Sarkar Prateek
Zhou Jiangying
Azarian Seyed
Harness & Dickey & Pierce P.L.C.
Johns Andrew W.
Matsushita Electric - Industrial Co., Ltd.
LandOfFree
Identification of logos from document images does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Identification of logos from document images, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Identification of logos from document images will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2592200