Image analysis – Image enhancement or restoration
Reexamination Certificate
1999-01-06
2002-07-16
Lee, Thomas D. (Department: 2624)
Image analysis
Image enhancement or restoration
C382S191000
Reexamination Certificate
active
06421468
ABSTRACT:
TECHNICAL FIELD
The present invention is related generally to digital image processing, and is related more particularly to image processing that enhances the sharpness of edges in a reproduction of an original image while minimizing or avoiding the emphasis of noise.
BACKGROUND ART
There are many types of processing of digital information representing images that can be used to bring about desirable changes in image reproductions generated from that information. For example, information representing colors or intensities in an image can be modified to achieve more realistic colors in a reproduction, can augment this information with additional information to increase the resolution or the spatial density of the elements making up a reproduction, can enlarge or shrink the size of a reproduction, or can combine information representing multiple images to produce a single composite image. Another example of image processing, referred to as image compression, reduces the amount of information used to convey an image. By reducing the amount of this digital information, an image can be conveyed by limited bandwidth transmission channels in less time, can be conveyed in a given amount of time over a transmission channel of lower bandwidth, or can be stored on media having reduced storage capacity.
Image compression techniques may be divided into two groups. In one group, “lossless compression” methods like Huffman coding and run-length coding reduce the entropy of the image information in a manner that can be reversed to recover the original image perfectly. In a second group, “lossy compression” methods like quantization reduce the capacity requirements in a manner that does not permit perfect recovery of the original image. Lossy compression methods tend to blur images and introduce noise. Preferably, lossy compression methods are applied selectively so these artifacts are not perceptible, or at least not objectionable, in the recovered image.
With the exception of lossless compression, the advantages offered by these types of processing are offset by degradations in the quality of the reproduction such as noise or reduced spatial detail. Because noise typically has a high spatial-frequency, many known techniques for reducing noise are essentially low-pass filtering processes. Unfortunately, low-pass filtering tends to reduce spatial detail, which is typically manifested by blurring and softening of edges.
Techniques for enhancing spatial detail, sometimes referred to as “image sharpening,” are essentially high-pass filtering processes. One well know method for image sharpening known as “unsharp masking” involves applying a low-pass spatial filter to an original image to obtain a blurred image with low spatial-detail. The difference between the blurred image and the original image, which is effectively a high-pass spatial filtered representation of the original image, is amplified by a scale factor and then added to the original image to obtain a sharpened image with emphasized high spatial-frequency components. Unfortunately, this process for image sharpening amplifies noise and sometimes sharpens edges or regions of an image that do not need sharpening. Furthermore, considerable memory is required to store the image so that the low-pass spatial filter can be applied with reasonable efficiency. Nevertheless, even with adequate memory, the filtering process is computationally intensive.
Several known techniques attempt to improve the basic unsharp masking technique by either improving the choice of scale factor used to amplify the blurred image, or by adapting the low-pass spatial filter used to obtain the blurred image. Unfortunately, these techniques do not reduce the filtering resources required to obtain the blurred image and, furthermore, require additional computational resources to compute scale factors or adapt filters.
In U.S. Pat. No. 5,038,388, incorporated herein by reference, a technique is described that attempts to optimize the amount of sharpening by computing a scale factor for each image element or “pixel” from a measure of pixel variances. The measure of pixel variances is obtained from two convolutional filter operations that are performed for each pixel. This technique requires additional resources to apply the convolutional filters.
In U.S. Pat. No. 5,363,209, incorporated herein by reference, a technique is described that adapts the frequency response of the spatial filter used to obtain the blurred image such that the maximum difference between nearby pixels is increased to a predetermined target value. This technique requires additional resources to determine local pixel differences and to adapt the spatial filter accordingly.
In U.S. Pat. No. 5,524,162, incorporated herein by reference, a technique is described that attempts to optimize the frequency response of a finite impulse response (FIR) spatial filter used to obtain the blurred image. This is accomplished by applying a Fast Fourier Transform (FFT) to the original image to determine the maximum spatial frequency f
MAX
in the image, and then calculating coefficients for the FIR spatial filter that increase the energy of the image components having a spatial frequency between 0.5 f
MAX
and f
MAX
. This technique requires additional resources to apply the FFT.
In U.S. Pat. No. 5,703,965, incorporated herein by reference, a technique is described that attempts to improve image sharpening by identifying edges in an original image prior to image compression, and then passing control information with the compressed image that describes the identified edges. This technique requires additional processing resources to identify edges and it reduces the effectiveness of image compression by requiring additional information capacity to convey the control information with the compressed imaged.
In U.S. Pat. No. 5,757,977, incorporated herein by reference, a technique applies complex filters that operate with fuzzy logic to discern edges from noise. In this manner, sharpening can be applied only to image features deemed to be edges. This technique requires additional resources for the logic to discern edges.
What is needed is a computationally efficient technique that can sharpen an image with proper consideration for noise and for edges that are already sufficiently sharp.
DISCLOSURE OF INVENTION
It is an object of the present invention to provide for a technique that selectively improves the sharpness of a reproduced image that requires only modest computational and memory resources.
According to one aspect of the present invention, a method or apparatus enhances the reproduction of an image by receiving a first signal and obtaining therefrom a first frequency-domain representation of the image, selecting one or more elements of the first frequency-domain representation according to one or more criteria, scaling the selected elements by a scale factor, forming a second frequency-domain representation of the enhanced reproduction of the image by combining the scaled selected elements with elements of the first frequency-domain representation that are not selected, and generating a second signal conveying the second frequency-domain representation.
According to another aspect of the present invention, a system for generating an enhanced reproduction of an original image comprises an input device and an output device in which the input device has first processing circuitry coupled to an input apparatus and coupled to a first terminal, wherein the first processing circuitry receives from the input apparatus an image signal representing the original image, transforms the image signal into a first frequency-domain representation of the original image, and generates at the first terminal a first signal that conveys the first frequency-domain representation; and in which the output device has second processing circuitry coupled to the first terminal and coupled to an output apparatus, wherein the second processing circuitry receives the first signal from the first terminal, obtains therefrom the first frequency-domai
Bhaskaran Vasudev
Ratnakar Viresh
Brinich Stephen
Gabrik Michael T.
Lee Thomas D.
Seiko Epson Corporation
LandOfFree
Method and apparatus for sharpening an image by scaling... 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 sharpening an image by scaling..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method and apparatus for sharpening an image by scaling... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2856413