Facsimile and static presentation processing – Facsimile – Recording apparatus
Patent
1988-01-28
1989-02-14
Groody, James J.
Facsimile and static presentation processing
Facsimile
Recording apparatus
358166, 358167, H04N 100, H04N 5213
Patent
active
048050318
DESCRIPTION:
BRIEF SUMMARY
The present invention relates to image processing methods and apparatus for reducing noise and increasing sharpness in an image recontructed from processed image signals generated by scanning and sampling an original image. More specifically, the invention pertains to such methods and apparatus in which sets of image element signals are processed using one-dimensional transforms applied to one-dimensional "windows" or image elements scanning the original image in a succession of various window orientations.
Photographic images, for example, are often impaired by unwanted non-image components such as graininess, commonly referred-to in the art of electronic image processing as "noise". To remove or, at least, reduce this unwanted component, various attempts have been made, one of which consists of dividing up the original image into image elements or "pixels", determining image values representative of the content of these pixels, processing these pixel image values so as to remove or reduce unwanted components in these values and then reconstructing a noise-free entire image from these "cleaned" image values. The processing of the image values broadly consists in replacing each pixel image value by a modified value, using the various image values within a field surrounding the pixel in order to form these modified values. The original pixel image values are used to form a number of different linear combinations, each of which represents a different component of the image structure within the field, and most of which represent components of the detail within the field. Each of these detail-sensitive combinations tends to vanish in the absence of a particular kind of image detail. Noise is reduced by modifying the detail sensitive combinations such that, for example, the value of a combination is lowered or set to zero whenever a particular kind of image detail is not present ot a significant degree. The modified pixel iage values are formed from these noise-reduced detail sensitive combinations.
One known way to build the above mentioned set of different linear combinations of pixel image values consists in applying to a set of values from (n.times.m) or "(n by m)", neighbouring pixels a mathematical (n.times.m) transform such as a Haar or Walsh-Hadamard transform. By dividing a full original image into adjacent blocks of (n.times.m) pixels and processing each block as per above, a full image can be processed and a noise-reduced image reconstructed.
Unfortunately, there is also a disadvantage associated with dividing the image into blocks. After processing, the reconstructed image may exhibit visible boundaries between adjacent blocks, giving a checkerboard appearance to that image, a feature unacceptable for high qualityimage reproduction.
U.S. Pat. No. 4,442,454 in the name of Philip G. Powell discloses an image processing method for reducing noise, which represents a substantial step forward in order to solve this difficulty. This method combines a block overlap transform process and a multistage procedure to overcome the problem of boundary visibility. According to that method, the image values within a small (n.times.m) "window" of pixels are transformed using a two-dimensional mathematical transform. The transformed image values are then thresholded to remove the small values that are primarily associated with noise, and the remainder are inverse transformed so as to regenerate a window of noise-reduced values. The window ismoved pixel by pixel, row-wise and column-wise, over the whole image, and the final value for each pixel is the average of (n.times.m) values derived from (n.times.m) overlapping windows. By this means the artefacts associated with fixed window boundaries are avoided (and those associated with thresholding the transformed values in each transform operation are reduced). According to the above-mentioned U.S. patent, the usefulness of this procedure is then increased by cascading several successive processing stages so as to build up larger windows fromarrays of smaller windows. Thes
REFERENCES:
patent: 4553165 (1985-11-01), Bayer
Eastman Kodak Company
Groody James J.
Owens Raymond L.
Vu Kim Y.
LandOfFree
Image processing method and apparatus using moving one-dimension does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Image processing method and apparatus using moving one-dimension, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Image processing method and apparatus using moving one-dimension will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-1370064