Image analysis – Image enhancement or restoration – Intensity – brightness – contrast – or shading correction
Reexamination Certificate
2000-10-06
2004-08-03
Mehta, Bhavesh M. (Department: 2625)
Image analysis
Image enhancement or restoration
Intensity, brightness, contrast, or shading correction
C382S167000, C382S266000, C382S199000, C358S519000
Reexamination Certificate
active
06771838
ABSTRACT:
FIELD OF THE INVENTION
The invention relates generally to the field of image processing, and more particularly to a system and method for enhancing digital images of documents.
BACKGROUND OF THE INVENTION
When documents are digitally captured in an uncontrolled environment such as in a casual setting, the illumination on the documents will most likely be non-uniform. As a result, the digital images of these documents may have an undesirable non-uniform illumination characteristic. For example, a document that is photographed by a digital camera may result in an image with regions that are unnaturally brighter or darker than other regions of the image.
One approach to correct the non-uniform illumination problem is to convert all pixel data of a digital image into binary data. Thus, the data from each pixel will be classified into one of two states, dark and light. If the value of a pixel is less than a prescribed threshold, the pixel is assumed to be a part of the text and is designated as a dark pixel. On the other hand, if the value of a pixel is greater than the prescribed threshold, the pixel is assumed to be a part of the background and is designated as a light pixel. In this fashion, any luminance variations in the digital image caused by non-uniform illumination can be eliminated.
A concern with the binary approach is that legitimate information on the digital image may be lost during the conversion. For example, if the original document includes pictorial regions, as well as text, the lighter areas of the pictorial regions may be designated as light pixel areas and be converted into background. Another concern is that grayscale and color information of the pictorial regions are lost during the binary conversion. Therefore, the binary approach is not well suited for document images containing pictorial regions.
Another approach to correct the non-uniform illumination problem is to use a homomorphic filtering technique to enhance the digital image. Using an illumination and reflectance model, this approach takes advantage of the fact that illumination components of a digital image, unlike the reflectance components, are generally characterized by slow spatial variations. Accordingly, the homomorphic filtering approach seeks to control the illumination components of the image by manipulating the low frequency components of Fourier transforms.
However, due to the multiplicative relationship between the illumination and reflectance components, these components cannot be directly converted into the frequency domain using Fourier transformation. This problem is resolved by taking the logarithm of the illumination and reflectance components and then performing a Fourier transform operation. In the frequency domain, the Fourier components that are associated with illumination are manipulated by using a homomorphic filter function to enhance the digital image. Two inverse operations, i.e., an inverse Fourier transformation and an exponentiation, are then sequentially performed to complete the process.
A concern with the homomophic filtering approach is that pictorial content may be damaged for a digital image containing both text and smooth pictorial regions, due to the linear and spatial-invariant filtering used to operate on the illumination and reflectance components.
In view of these concerns, there is a need for a system and method to correct the effect of non-uniform illumination in digital images containing both text and pictorial content without degrading the pictorial content.
SUMMARY OF THE INVENTION
A system and method for enhancing digital images containing both text and pictorial content (“mixed document images”) utilizes an estimated illumination surface of a given digital image to correct the undesirable effect of non-uniform illumination. The use of an estimated illumination surface allows the system and method to correct the effect of non-uniform illumination for mixed document images without loss of grayscale or color information, even for mixed document images containing large smooth pictorial regions. In addition to non-uniform illumination correction, the system and method enhances the mixed document images by sharpening and/or darkening edges of text contained in the images.
In an exemplary embodiment, the system in accordance with the invention includes an edge detector, a text enhancer, an illumination surface estimator, and an illumination corrector. These components may be implemented in any combination of hardware, firmware and software. In one embodiment, these components are embodied in the system as a software program that is executable by a processor of the system.
The edge detector operates to detect edges of text contained in a given digital image. The text enhancer operates to sharpen and darken the detected text edges. The illumination surface estimator operates to generate an estimated illumination surface of the digital image. The estimated illumination surface is based on the luminance values of the edge pixels that are on the dark side of the detected text edges. In an alternative embodiment, the luminance values of edge pixels that are on the lighter side of the detected text edges are used to generate the estimated illumination surface. The illumination corrector of the system operates to correct the effect of non-illumination by adjusting pixel values of the digital image using the estimated illumination surface. The illumination correction is accomplished by dividing the pixel values of the digital image by corresponding values of the estimated illumination surface.
In an exemplary embodiment, the method in accordance with the invention includes the step of receiving a digital image having a non-uniform illumination characteristic. The method also includes the step of detecting edges of text contained in the digital image and the step of enhancing the text using text edge information. The text enhancement may include sharpening and darkening the text edges. The method further includes the step of generating an estimated illumination surface of the digital image and the step of selectively adjusting pixel values of the digital image using the estimated illumination surface to correct the effect of non-uniform illumination.
Other aspects and advantages of the present invention will become apparent from the following detailed description, taken in conjunction with the accompanying drawings, illustrated by way of example of the principles of the invention.
REFERENCES:
patent: 5414538 (1995-05-01), Eschbach
patent: 6115022 (2000-09-01), Mayer et al.
patent: 6462768 (2002-10-01), Oakley
patent: 2002/0025079 (2002-02-01), Kuwata et al.
Rafael C. Gonzalez and Paul Wintz, Digital Image Processing, second edition, Addison-Weslye Publishing Company, Reading, Massachusetts, 1987, pp. 183-186.
Mohamed Kamel and Aiguo Zhao, “Extraction of Binary Character/Graphics Images from Grayscale Document Images”, CVGIP, vol. 55, No. 3, 1993, pp. 203-217.
William H. Press, Saul A. Teukolsky, William T. Vettering and Brian P. Flannery, Numerical Recipes in C, second edition, Cambridge University Press, New York, 1992, pp. 123-124.
Hewlett--Packard Development Company, L.P.
Mehta Bhavesh M.
Strege John
LandOfFree
System and method for enhancing 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 System and method for enhancing document images, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and System and method for enhancing document images will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3271700