Image analysis – Image enhancement or restoration – Image filter
Reexamination Certificate
1997-03-31
2002-09-17
Boudreau, Leo (Department: 2621)
Image analysis
Image enhancement or restoration
Image filter
C382S162000, C382S299000
Reexamination Certificate
active
06453074
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates to computerized image and color image processing, and more particularly to image decimation with adaptive pre-filtering whereby a pixel-by-pixel determination is made of image busyness based on which it is decided whether to pre-filter before decimation.
2. Description of the Related Art
As resolution of computerized images and color images increases, the amount of storage needed to store such images has increased dramatically. For example, for computerized images at 400 dots per inch (dpi) resolution, each square inch of image area is formed by a 400×400 pixel matrix. Consequently, for a standard 8½×11 inch page, information for 14,960,000 pixels is needed. For a full color image formed by three color components (such as red, green and blue (RGB) color components), at eight bits per color component, 45 megabytes of memory are needed.
Processing 45 megabytes of information strains even the most advanced personal computing equipment available today. For example, simply to display 45 megabytes of a color image strains the bandwidth available on a typical computer bus. Accordingly, a long time is needed simply to display the image.
Furthermore, once the image is displayed, it is commonplace to manipulate the image such as by adjusting color and contrast, or by applying various image processing operations such as sharpness enhancements, posterization, special effects, and the like. Such image manipulations require a long time when applied to a full 45 megabyte image. Moreover, it is commonplace for such manipulations to be arrived at through a process of trial and error by which various image processing operations are tried until a pleasing effect is arrived at. When these-trial and error processes are applied to a full 45 megabyte image, the computational power of even advanced personal computing equipment severely limits the speed at which a pleasing effect can be determined.
In consideration of the foregoing, it has been proposed to decimate the high resolution image prior to display and manipulation, and to use the decimated image as a proxy both for display and manipulation. A decimated image is a low resolution image derived from an original high resolution image, and is typically derived by sub-sampling the image, such as by selecting every other pixel in both horizontal and vertical directions. Once a decimated image has been derived, it is the decimated image that is displayed, and it is the decimated image upon which the trial-and-error process of determining pleasing image manipulations is effected. Once a pleasing effect has been determined, it is applied to the high resolution image. Thus, the decimated image serves as a proxy for the high resolution image, permitting rapid display and speedier derivation of pleasing visual effects.
Derivation of a decimated image proceeds in two steps: a first step in which the high resolution image is subjected to low-pass filtering, and a second step in which the filtered image is sub-sampled. Low-pass filtering is applied so as to avoid the introduction of artifacts in the decimated image by eliminating high frequency components that might cause signal aliasing during sub-sampling. While such low-pass filtering is necessary at high frequency areas of the high resolution image, unwanted side effects are often introduced. For example, low-pass filtering causes blurring of the image, and can introduce a color shift when the frequency content of one color component differs significantly from the other two.
Thus, while low-pass filtering prior to sub-sampling is needed to prevent signal aliasing caused by sub-sampling, such low-pass filtering introduces unwanted side effects that have not heretofore been adequately addressed in the art.
SUMMARY OF THE INVENTION
It is an object of the present invention to address the disadvantages found in conventional image decimation processes which involve low-pass filtering followed by sub-sampling, and to address these disadvantages through adaptive pre-filtering in which the application of a low-pass filter to a target pixel depends on the local busyness of the image.
Thus, in one aspect, the invention is image decimation in which a pixel-by-pixel determination is made of busyness in the image, based on which it is decided whether or not to low-pass filter before sub-sampling. As used herein, “busyness” refers to the spatial frequency content of a neighborhood of pixels surrounding a target pixel. The spatial frequency content can be determined by standard transformation techniques such as Fourier or Hough transformation, or it may be made through a simple comparison of intensity values of the target pixel to those of adjacent surrounding pixels. Based on the determination of busyness, the characteristics of low-pass filtering that is applied to the target pixel are determined. For example, based on the determination of busyness, it might be decided not to apply any low-pass filtering for a pixel with low busyness, whereas low-pass filtering would be applied to a pixel that does not have low local busyness. As another example, different degrees of low-pass filtering might be applied based on the amount of local busyness.
For black-and-white or grayscale images in which there is only one color component (or in which all three of the RGB color components are equal), busyness can be determined based on only a single color component and adaptive filtering applied to the same color component. For color images, on the other hand, it is preferable to transform the color image into a color space that closely matches human visual perception if the color image is not already in such a color space. For example, for an RGB color image, it is preferable to transform the image into CIEL*a*b* color space prior to a determination of busyness. Thereafter, the determination of busyness is made based only on the L* component. Equal adaptive pre-filtering may thereafter be applied to all three color components in the event that local busyness of the L* component indicates that such pre-filtering is desirable, but it is more preferable to pre-filter only the L* component since the a* and b* components are responsible primarily only for chromaticity. Thereafter, the adaptively pre-filtered L*a*b* is transformed back into RGB color space, for sub-sampling to form the decimated image.
Good results of color images are also obtained without transforming to different color spaces. In this case, it is possible to treat each color component independently, with each color component being adaptively pre-filtered and sub-sampled independently of the other two color components. Likewise, it is also possible to select one color component such as green (which is known to closely follow the brightness component in YIQ color space) as the color component for which a busyness determination is made, with all three color components being adaptively pre-filtered based on the determination of busyness for the selected single color component.
This brief summary has been provided so that the nature of the invention may be understood quickly. A more complete understanding of the invention can be obtained by reference to the following detailed description of the preferred embodiment thereof in connection with the attached drawings.
REFERENCES:
patent: 4561022 (1985-12-01), Bayer
patent: 4691233 (1987-09-01), Acampora
patent: 4868653 (1989-09-01), Golin et al.
patent: 4918523 (1990-04-01), Simon et al.
patent: 4974065 (1990-11-01), Murakami et al.
patent: 5047838 (1991-09-01), Murakami et al.
patent: 5384869 (1995-01-01), Wilkinson
patent: 5450216 (1995-09-01), Kasson
patent: 5566284 (1996-10-01), Wakayama
patent: 5778104 (1998-07-01), Kowaski
patent: 5845017 (1998-12-01), Keyes
patent: 5933542 (1999-08-01), Chang et al.
Boudreau Leo
Sherali Ishrat
LandOfFree
System for image decimation including selective filtering 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 for image decimation including selective filtering, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and System for image decimation including selective filtering will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2823672