Image analysis – Image enhancement or restoration – Variable threshold – gain – or slice level
Reexamination Certificate
1999-12-22
2004-10-12
Werner, Brian (Department: 2621)
Image analysis
Image enhancement or restoration
Variable threshold, gain, or slice level
C382S274000, C382S275000
Reexamination Certificate
active
06804408
ABSTRACT:
FIELD OF THE INVENTION
The invention relates generally to the field of digital image processing and, more particularly, to a method for enhancing the texture of a digital image.
BACKGROUND OF THE INVENTION
Traditional methods of increasing the apparent sharpness of a digital image, such as the technique of unsharp masking, often produce unwanted artifacts at large transition edges in the image. For example, unsharp masking is often described by the equation:
Sproc=Sorg+B
(
Sorg−Sus
)
where Sproc represents the processed image signal in which the high frequency components have been amplified, Sorg represents the original image signal, Sus represents the unsharp image signal, typically a smoothed image signal obtained by filtering the original image, and B represents the high frequency emphasis coefficient.
The unsharp masking operation may be modeled as a linear system. Thus, the magnitude of any frequency in Sproc is directly dependent upon the magnitude of that frequency in the Sorg image signal. As a consequence of this superposition principle, large edges in the Sorg image signal will often display a ringing artifact in the Sproc signal when the desired level of high frequency enhancement has been performed in other areas of the Sproc signal. This ringing artifact appears as a light or dark outline around the large edge, and may be visually objectionable.
Many non-linear filters based on local statistics exist for the purposes of noise reduction, sharpening, and contrast adjustment. For example, the median filter is well known in the art. In this filter, typically implemented for noise reduction, each pixel is replaced with the median value of some surrounding neighborhood. This filtering process is generally very successful at removing impulse noise; however, the processed image appears slightly less sharp than the original.
Another example of a non-linear filter based on local statistics is local histogram equalization, referred to as adaptive histogram modification by William Pratt on pages 278-284 of the book
Digital Image Processing, Second Edition
, John Wiley & Sons, 1991. With this filter, the values of pixels are modified by the cumulative histogram of a local window. This technique effectively adjusts the contrast of each region of a digital image, effectively increasing the local contrast in some regions of the image, and decreasing the contrast in other regions. This technique does not intend to increase the apparent sharpness of any given region. Also, this technique does not ensure that the typical artifacts of ringing will not occur.
There exist many algorithms for sharpening the appearance of images without generating artifacts or enhancing the notability of noise. In U.S. Pat. No. 4,571,635, Mahmoodi and Nelson teach the method of deriving an emphasis coefficient B that is used to scale the high frequency information of the digital image depending upon the standard deviation of image pixel values in a local neighborhood. In addition, in U.S. Pat. No. 5,081,692, Kwon and Liang teach that the emphasis coefficient B is based upon a center weighted variance calculation. However, neither Mahmoodi et al nor Kwon et al consider the expected standard deviation of noise inherent in the imaging system. By not considering the noise inherent in the imaging system, both Mahmoodi and Kwon make the implicit assumption that all imaging sources and intensities have the same noise characteristics. In addition, neither use separate strategies for texture and edge regions.
In U.S. Pat. No. 4,794,531, Morishita et al teaches a method of generating the unsharp image with a filter whose weights on neighboring pixels are based upon the absolute difference between the central pixel and the neighboring pixel. Morishita claims that this method effectively reduces artifacts seen at the edges of a sharpened image (as compared with traditional unsharp masking). In addition, Morishita sets a gain parameter based upon local standard deviation and the standard deviation of the entire image. Again, Morishita does not consider the levels of noise inherent in the imaging system in order to approximate signal to noise ratios. In addition, the method of Morishita does not offer explicit control over the amount of edge reshaping.
In U.S. Pat. No. 5,038,388, Song teaches a method of amplifying image details without amplifying the image noise by adaptively amplifying the high-frequency component of the image. An estimate of image noise power is used, however; this noise power is not described as being dependent on the intensity or the pixel. In addition, Song does not attempt to estimate signal to noise ratio in order to control the level of sharpening.
Thus, there exists a need for an alternative method of manipulating a digital image in order to generate an image signal that appears to be sharper, or more in focus, while minimizing the ringing artifact that is evident with the unsharp masking technique and enhancing the magnitude of detail in the scene in a noise sensitive manner.
SUMMARY OF THE INVENTION
It is an object of this invention to allow for independent control of the enhancement applied to detail, large edges, and noisy regions.
The present invention is directed to overcoming one or more of the problems set forth above. Briefly summarized, according to one aspect of the present invention, the invention resides in a method for enhancing a digital image channel, e.g., a channel comprising a texture signal, by providing a predetermined estimate of the noise expected for the digital channel based on a predetermined relationship between the image intensity values and the expected noise for given intensities. After a local estimate of signal activity is generated for the digital image channel, a gain adjustment is generated from the predetermined estimate of noise and the local estimate of signal activity; and the gain adjustment is applied to the image pixels in the digital channel in order to generate a digital channel with enhanced image values.
The present invention has the advantage of boosting the texture signal by a factor that is related to an estimate of the local signal to noise (SNR) ratio. Thus, the portion of the texture signal coincident with regions of the digital image channel having high SNR, such as the modulation of a grassy field due to many blades of grass in addition to image system noise, will experience a greater level of boost as compared with the portion of the texture signal associated with low SNR regions of the digital image channel, such as a large region of clear blue sky where the only modulation is likely to be noise resulting from the imaging system. Therefore, while it is not desirable to increase the amplitude of regions having only noise modulation, it is preferable to boost the modulation where it can be attributed to actual modulation in the scene. In the present invention, the signal to noise ratio functions as a classifier that may be used to distinguish between, e.g., the two aforementioned types of regions in a scene.
These and other aspects, objects, features and advantages of the present invention will be more clearly understood and appreciated from a review of the following detailed description of the preferred embodiments and appended claims, and by reference to the accompanying drawings.
REFERENCES:
patent: 4591923 (1986-05-01), Urabe et al.
patent: 4794531 (1988-12-01), Morishita et al.
patent: 4945502 (1990-07-01), Kwon et al.
patent: 5038388 (1991-08-01), Song
patent: 5081692 (1992-01-01), Kwon et al.
patent: 5111285 (1992-05-01), Fujita et al.
patent: 5134503 (1992-07-01), Kimura
patent: 5641596 (1997-06-01), Gray et al.
patent: 5923775 (1999-07-01), Snyder et al.
patent: 6295382 (2001-09-01), Karanovic
patent: 0 851 666 (1998-07-01), None
patent: 0 851 666 (1998-07-01), None
U.S. patent application Ser. No. 09/324,239, Gallagher et al., filed Jun. 2, 1999.
“Rational unsharp masking technique” by Giovanni Ramponi and Andrea Polesei. Journal of Electronic Imaging, Apr. 1998, vol. 7(2), pp. 333-338.
Gallagher Andrew C.
Gindele Edward B.
Eastman Kodak Company
Werner Brian
Woods David M.
LandOfFree
Method for enhancing a digital image with noise-dependent... 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 for enhancing a digital image with noise-dependent..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method for enhancing a digital image with noise-dependent... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3282805