Image analysis – Color image processing
Reexamination Certificate
1999-05-27
2001-05-01
Tran, Phuoc (Department: 2621)
Image analysis
Color image processing
C382S165000, C382S190000
Reexamination Certificate
active
06226399
ABSTRACT:
TECHNICAL FIELD
This invention relates to methods and systems for identifying an image feature and methods and systems for determining an optimal color space for use therein.
BACKGROUND ART
A very important image processing tool used in machine vision automatically identifies given features within the image. Up to recently, this has almost always been accomplished using a monochrome system in conjunction with some form of search algorithm. However, there are situations where there is insufficient grey level difference between the intended target and the background for the target to be reliably identified in this manner. In many cases, switching from a monochrome imaging system to a color system can be of great assistance, relying on using color discrimination to separate the target from the background.
U.S. Pat. No. 5,751,450 discloses a tool which has been designed to ease the construction of such color based discriminators using what is, in effect, a programmable color filter. However, there is an important class of images in which the color difference is insufficient for the straight forward color filter to work effectively.
Conventionally, color images are captured and manipulated in a color space defined by the red, green, and blue axes, as illustrated in FIG.
2
. This is primarily because both the human eye, and virtually all color cameras come equipped with sensors that are sensitive to these colors, and so it becomes second nature to operate in such a color space, or one closely related to it. However, in many images this choice of axes could be far from optimum, especially if there is any information common to the red, green, and blue axes.
An example of the worse case of such a distribution is shown in FIG.
2
. Here, every pixel in a sample image is plotted in its correct position in the RGB color space. The result, in this case, has a very strong correlation to a straight line that is at an angle of 45° to all of the axes. With this type of distribution, the image information is effectively shared equally over the three color planes. Although this represents the most unfavorable type of distribution, it is not particularly uncommon - occurring anywhere that the predominant color is grey, for example imaging aluminum or stainless steel structures.
U.S. Pat. No. 4,653,014 to Mitcami et al. discloses a process for preparing discrimination criteria for identifying colors in a color identifying system. A computer-based system attempts to automatically determine the best way in which to be able to distinguish between single color samples. The system must be shown the color which it is trying to identify. It will then automatically derive the optimum parameters with which to be able to differentiate samples of these colors. It describes a spot measuring instrument rather than an area measuring instrument. Its output is essentially digital, inasmuch as the output result is that a sample is identified as being identical to one of the original samples. There is no way to quantitatively measuring how close, or how far the sample is from the original reference.
U.S. Pat. No. 5,218,555 to Komai et al. discloses a method for judging a color difference using rules and fuzzy inference and apparatus therefor. The color matching system is totally based on “fuzzy logic” and “fuzzy inference” components within its algorithm.
U.S. Pat. No. 5,410,637 to Kern et al. discloses a color tolerancing system employing fuzzy logic. This technique appears to be designed for matching a single color sample. As with the Komai et al. patent, the algorithm is totally reliant on “fuzzy logic”.
U.S. Pat. No. 5,085,325 to Jones et al. discloses a color sorting system and method. The system uses a polar coordinate color specification. An external look up table is used to perform an acceptance/rejection flag. Digitized video goes straight into the hardware-implemented look up table for real-time speed.
U.S. Pat. No. 5,221,959 to Ohyama et al. discloses a color discrimination data input apparatus. The system modifies the spectral content of the illumination on an object in order to obtain the optimum separation of several classes of object based on differences in spectral reflectance of the objects being classified.
U.S. Pat. No. 4,414,635 to Gast et al. discloses a system for segmenting an image based on certain key hues. The system is trained on a series of keystone colors, each of which is given a label. These colors only cover a small segment of the color space, thereby leaving many colors unlabeled. In order to ensure that every color has a label, Gast uses color distance to determine which key color any unlabeled hue is closest to and allows that color to adopt the label of its neighboring key color. From this process, Gast generates a look up table whose address defines the color of the current pixel and the output generates the label. By scanning another image pixel by pixel, each part of the new image can be allocated a label. Gast extends the system to produce a complementary facility whereby labels can be converted back to the key color. Thus, by using a pre-trained segmenter to scan an image followed by a synchronized complementary system, Gast can generate a duplicate image which only contains the key hues.
U.S. Pat. No. 5,552,805 to Alpher is also related to the present invention.
DISCLOSURE OF INVENTION
An object of the present invention is to provide a method and system for identifying an image feature and method and system for determining an optimal color space for use therein wherein image features can be identified even though only small color differences exist between the image features and background of the image features.
Another object of the present invention is to provide a method and system for identifying an image feature and method and system for determining an optimal color space for use therein wherein image features of interest are enhanced by using a color axis rotation in an optimal direction.
In carrying out the above objects and other objects of the present invention, a method is provided for identifying an image feature of an object in a machine vision system. The method includes the step of capturing an original color image of the object in a first color space defined by a first set of color axes to obtain a first set of image data. The method also includes the step of transforming the first set of image data to a second set of image data in an optimal color space having an optimal set of color axes which define an optimal direction wherein the amount of inter axis correlation of the second set of image data is minimized. The method further includes the step of manipulating the second set of image data to identify the image feature wherein color differences existing in the original color image are amplified and separation between chromatically different regions of the original color image are maximized in the optimal color space.
Preferably, the optimal set of color axes are three orthogonal axes.
Typically, the step of transforming the first set of image data includes the step of chromatically rotating the first set of image data to match the optimal direction. In other words, the step of transforming the first set of image data includes the step of reorganizing the first set of image data along the optimal set of color axes.
The method may also include the steps of capturing a color image of an ideal object in the first color space to obtain an ideal set of image data and computing the optimal set of color axes from the ideal set of image data wherein the optimal set of color axes is object specific.
Further in carrying out the above objects and other objects of the present invention, a method is provided for determining an optimal color space defined by an optimal set of color axes for an object in a machine vision system. The method includes the steps of capturing a color image of an ideal object in a first color space having a first set of color axes to obtain an ideal set of image data and computing an optimal set of axes which define the optimal color space fro
Brooks & Kushman P.C.
Integral Vision, Inc.
Tran Phuoc
LandOfFree
Method and system for identifying an image feature and... 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 system for identifying an image feature and..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method and system for identifying an image feature and... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2500563