Image analysis – Color image processing
Reexamination Certificate
1998-02-27
2001-03-27
Couso, Yon J. (Department: 2723)
Image analysis
Color image processing
C382S167000
Reexamination Certificate
active
06208753
ABSTRACT:
FIELD OF THE INVENTION
This application relates to the field of digitized imaging and is more specifically directed to improving scanned image quality.
BACKGROUND OF THE INVENTION
The process of scanning and digitizing a physical scene to form a digitized color image begins by optically focusing the physical scene onto an imaging plane. The imaging plane is spatially divided into a two-dimensional array of small, uniform elements called pixels. The color scanning process typically involves a scanning device that performs several passes of a light sensitive device over the imaging plane to measure the average intensity of irradiated light at each pixel location. On each pass, the light sensitive device is covered by a different colored filter having a carefully chosen spectral transmittance. At least two passes are typically involved. Each pass of the light sensitive device over the imaging plane produces a color plane. Associated with each pixel in each color plane is an integer called a color component. The color component is used to record the average value of the digitized intensity of irradiated light measured by the light sensitive device. The plurality of integers in corresponding positions in the several color planes make up the family of color components of the pixel. The plurality of pixel color components for all color planes make up the digitized image.
Some present-day image scanning devices are capable of measuring and recording the integer values of all color components of each pixel during a single pass over the imaging plane. Still other devices are capable of measuring and recording the values of all color components of all pixels simultaneously. The net result, however, is the same. Each device produces the values for the several color planes that represent the digitized color image.
In what follows, whenever reference is made to color planes, it is understood to include any number of color planes used by a particular image digitizing technique to define pixel color components.
A digitized image is recognizable as an image by a human viewer only when the individual pixel color components are displayed as dots of colored light on a display, or as dots of colored inks or dyes on a hardcopy. Pixels are normally chosen to be spaced so closely as to be unresolvable by a human visual system. This results in the fusion of neighboring pixels by the human visual system into a representation of the original physical scene.
The optical paths between the physical scene and the light sensitive device are different for each of the color planes.
This is because a different colored filter is used for each color plane. Each of these filters has minor variations in its thickness, index or refraction, homogeneity, and in its physical placement and alignment. These variations can not be entirely eliminated with the p resent state of the art. In addition, minor mechanical variations occur in scanning devices that typically involve physical movement of either the scene or the light sensitive device, or both. These minor mechanical variations can not be entirely eliminated with the present state of the art. Any or all of these variations, although minimized by careful manufacture of the scanner, cause misregistration among the several color planes. Thus features of an object in the physical scene appear in slightly different pixel positions in each of the several color planes. Misregistration of the color planes produces the sensation of sub-optimal focus to a human viewer when the image is displayed at its full resolution. It also produces “false-color” fringes at the edges of features in the image where an abrupt color change occurs. This is especially evident when the digitized image is enlarged.
A detailed description of color component values is found in G. Wyszecki and W. S. Styles, “Color Science: Concepts and Methods, Quantitative Data and Formulae,” John Wiley & Sons, Inc. (2nd ed.), New York, 1982, pp. 164-169, incorporated here in by reference in its entirety. The CIE 1931 standard specifies three particular reference stimuli. The stimuli are radiometric quantities, and as such are expressed in radiometric units such as watts. Grassmann's law, on which nearly all of modern colorimetry is based, requires use of those three specific reference stimuli, or three others that are distinct linear combinations of them. This is discussed in D. B. Judd and G. Wyszecki, “Color in Business, Science, and Industry,” (3rd ed.), John Wiley & Sons, Inc., New York, 1975, pp. 45-47, incorporated herein by reference in its entirety. Also incorporated herein by reference in its entirety, is a paper by F. Mintzer and G. W. Braudaway, “Processing Color Images while Preserving Their Color,” Proceedings of the Ninth Workshop on Image and Multidimensional Signal Processing, Mar. 3-6, 1996, Belize City, Belize, pp. 106-107.
SUMMARY OF THE INVENTION
The present invention provides an article of manufacture, system and process to be applied to the several color planes representing an image, by which misregistration among the pixel components in the various color planes can be measured and, in large part, removed. This brings the several planes into more substantially perfect registration. This process is herein called reregistration.
In one aspect the present invention provides a process by which misregistration among the pixel components of the several color planes of a color image can be substantially eliminated. The process begins with the scanning of a test scene. The physical object which comprises the test scene typically has features that are uniform squares of alternating lack and white color arranged, for instance, in a grid forming a black and white checkerboard. Other test scene grids ay be used. The digitized color image of the test scene is employed to make measurements of misregistration of scene features in its color planes.
In an example embodiment, a process by which misregistration among the several color planes is measured, and substantially eliminated, is made up of three parts, described as follows:
Firstly, one of the color planes is chosen as a reference plane. The remainder of the color planes are denoted secondary planes. The digitized image must have at least two color planes, that is, one reference plane and at least one secondary plane. The relative misregistration of image features in each secondary plane relative to corresponding image features in the reference plane is measured at several reference points at dispersed pixel locations in the reference plane. Typically, four or more pixel locations are used. Each measured misregistration includes a two-dimensional offset of an image feature in a secondary plane with respect to the corresponding feature in the reference plane. One type of two-dimensional offset is a horizontal and vertical offset.
Secondly, for each of the secondary planes, an interpolated approximation of the horizontal and vertical offset of scene features is computed for every pixel location in the reference plane by using the vertical and horizontal offsets measured at the four or more dispersed pixel locations,
Thirdly, using the approximate feature offsets computed as above, the color components of each pixel in each of the secondary planes can be interpolated using those offsets and the color component values from the neighboring pixels in the same secondary color plane.
The interpolated color component values derived from each of the secondary planes are used to form revised color planes. The revised planes are used in place of the secondary planes and in conjunction with the reference plane to form the reregistered image.
The small systematic imperfections inherent in even the best scanners tend to produce very repeatable misregistrations. Thus, once a correction for misregistration has been computed for the test scene, it can be applied in an identical manner to any other physical scene that is focused in a like manner on the imaging plane of a particular scanner. In this sense, the image of the test scene and its reregistration parameters
Braudaway Gordon Wesley
Christensen James Edward
Thompson Gerhard Robert
Couso Yon J.
Herzberg Louis P.
International Business Machines - Corporation
LandOfFree
Quality of digitized images through post-scanning... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Quality of digitized images through post-scanning..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Quality of digitized images through post-scanning... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2511029