Image analysis – Image segmentation – Distinguishing text from other regions
Reexamination Certificate
1998-10-22
2001-09-25
Chang, Jon (Department: 2623)
Image analysis
Image segmentation
Distinguishing text from other regions
C358S462000, C358S464000
Reexamination Certificate
active
06295371
ABSTRACT:
BACKGROUND OF THE INVENTION
This invention relates to a method for image processing employing image segmentation using tokenization. More particularly, the present invention is directed to a method for segmenting document images into text (or symbols) and continuous tone regions using a shape comparison, or tokenization, as an indicator of which high contrast regions of the image correspond to text. Such segmentation is useful for implementation of image processing techniques such as compression and decompression.
While the invention is particularly directed to the art of image processing including segmentation of images, and will thus be described with specific reference thereto, it will be appreciated that the invention may have usefulness in other fields and applications.
By way of background, image processing has gained considerable recognition in recent years. Unfortunately, however, it is still difficult for a computer to “look at” a scanned, color image and determine which parts of that image correspond to text and which parts correspond to continuous tone regions.
This is significant because segmenting images into text and continuous tone regions is useful from an image processing standpoint. Once text is separated out from the image, it can be more efficiently processed by way of Optical Character Recognition (OCR), for example, apart from the continuous tone regions.
In addition, where image compression and decompression are implemented, segmentation of the text from other parts of the image is important because continuous tone images can be stored at a lower resolution than text without visible degradation. Most known continuous tone, lossy compression methods result in blurred text because these compression methods do not effectively deal with high contrast regions of text.
Therefore, it would be extremely valuable to have an effective segmentation method that would determine parts of images that correspond to text and parts that do not. Known methods are deficient.
In this regard, known methods for segmenting images typically use pixel level statistics. That is, these methods consider an area or region of an image and make determinations based on contrast between pixels therein. For example, one of these known methods may determine that a high contrast area in a particular region being analyzed corresponds to text; however, such a method does not effectively deal with the situation where nontextual high contrast regions are present in the image. Therefore, images that are not text will be improperly treated as such, resulting in false positive results.
Methods have been proposed to compensate for these false positive results when using pixel level statistics on a local basis. However, when these methods are employed, certain representations of text get lost because of overcompensation in narrowing the criteria for determination of whether a component is text.
In addition, known methods concentrate only on intensities of pixels. These methods do not consider shape as part of the decision as to whether a part of an image is text or a continuous tone region.
The present invention provides a new and improved image processing method which overcomes the above noted problems and difficulties.
SUMMARY OF THE INVENTION
A method is provided for image processing employing a process for segmenting an image that is comprised of text and continuous tone regions. Shapes of components are used as criteria for determining text in the segmentation process.
In one aspect of the invention, a thresholding routine is performed on the image to transform the color image to a black and white image including black and white components, selecting a group of components based on predetermined criteria such as color variance, Hamming distances and alignment, classifying the components of the selected group based on shape and identifying certain components as text based on the classification (i.e. shape recurrence), color, alignment, proximity and relative size.
In another aspect of the invention, the identified text is removed from the image to obtain a background image.
In another aspect of the invention, the identified text is processed.
In another aspect of the invention, the background image is processed apart from the text.
Further scope of the applicability of the present invention will become apparent from the detailed description provided below. It should be understood, however, that the detailed description and specific examples, while indicating preferred embodiments of the invention, are given by way of illustration only, since various changes and modifications within the spirit and scope of the invention will become apparent to those skilled in the art.
REFERENCES:
patent: 4466121 (1984-08-01), Damen et al.
patent: 5539841 (1996-07-01), Huttenlocher et al.
patent: 5835638 (1998-11-01), Rucklidge et al.
patent: 5848185 (1998-12-01), Koga et al.
patent: 5930393 (1999-07-01), Ho et al.
patent: 6026182 (2000-02-01), Lee et al.
patent: 6035059 (2000-03-01), Kurosawa et al.
Luc Vincent, “Morphological Algorithms”, Harvard Robotics Laboratory, Technical Report No. 91-12, pp. 1-30.
Felzenswalb Pedro
Huttenlocher Daniel P.
Rucklidge William J.
Chang Jon
Fay Sharpe Fagan Minnich & McKee LLP
Wu Jingge
Xerox Corporation
LandOfFree
Method and apparatus for image processing employing image... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Method and apparatus for image processing employing image..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method and apparatus for image processing employing image... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2455076