Image analysis – Color image processing – Pattern recognition or classification using color
Reexamination Certificate
1999-08-17
2004-06-29
Do, Anh Hong (Department: 2624)
Image analysis
Color image processing
Pattern recognition or classification using color
C382S162000, C382S164000, C382S209000, C382S219000
Reexamination Certificate
active
06757428
ABSTRACT:
FIELD OF THE INVENTION
The present invention relates to characterizing colors in an image, as well as to measuring colors in an image and matching colors between different image objects in a template and a target image.
DESCRIPTION OF THE RELATED ART
In machine vision applications, color is a powerful descriptor that often simplifies object identification and extraction from a scene. Color characterization, location, and comparison is an important part of machine vision and is used in a large class of assembly and packaging inspection applications. Inspection involves verifying that the correct components are present in the correct locations. For example, color may be used to inspect printed circuit boards containing a variety of components; including diodes, resistors, integrated circuits, and capacitors. These components are usually placed on a circuit board using automatic equipment, and a machine vision system is useful to verify that all components have been placed in the appropriate positions. A method of effectively characterizing and locating components on a circuit board is desired.
Color is widely used in the automotive industry to verify the presence of correct components in automotive assemblies. Components in these assemblies are very often multicolored. For examples, color characterization may be used to characterize and inspect fuses in junction boxes, i.e. determining if all fuses are present and in the correct locations. In addition, it is often necessary to match a fabric in one part of a multi-color automobile interior. A method of characterizing a multi-colored area for analysis and comparison is desirable. A color characterization may be used to determine which of several fabrics is being used as well as the orientation (texture) of the fabric. In a manufacturing environment, color matching techniques may also be used to find flaws in a manufactured product, assuming the flaws are accompanied by a color change.
A direct measurement of the color distribution in an object may also be useful in machine vision applications. Instead of matching colors between the target image and the template image, a quantitative representation of the color distribution is produced. This data may then be used for further manipulation of the color information in machine vision applications. For example, a color measurement machine vision system may be used to test and analyze the color components in textiles or dyes. A quantitative representation is also useful in color image segmentation and color image retrieval. For example, an object in an image may be segmented based on its color, and an image may be retrieved from an image database based on its color representation.
A color space (or color model) is a way of representing colors and their relationship to each other. A color space is essentially a 3-D coordinate system and a subspace within that system where each color is represented by a single point or vector. Image processing and machine vision systems use several different color spaces including RGB, HSI (or HSL) and CMY. In the RGB space, each color appears in its primary spectral components of red, green and blue. This RGB color space is based on a Cartesian coordinate system. The RGB model is represented by a 3-dimensional cube with red, green, and blue at the edges of each axis. Each point in the cube represents a color, and the coordinates of that point represents the amount of red, green and blue components present in that color. Because the red, green, and blue color components in RGB color space are highly correlated, it is difficult to characterize colors with intensity/luminance independent features.
The Hue, Saturation, Intensity (HSI) or Hue, Saturation, Luminance (HSL) color space was developed to put color in terms that are easier for humans to quantify. The hue component is color as we normally think; such as orange, green, violet, and so on (a rainbow is a way of visualizing the range of hues). Thus, hue represents the dominant color as perceived by an observer. Saturation refers to the amount or richness of color present. Saturation is measured by the amount of white light mixed with a hue. In a pure spectrum, colors are fully saturated. Colors such as pink (red and white) and lavender (purple and white) are less saturated. The intensity or light component refers to the amount of grayness present in the image.
Colors represented in HSI model space may be ideal for machine vision applications for two reasons. First, HSI includes an intensity (luminance) component separated from the color information. Second, the intimate relation between hue and saturation more closely represents how humans perceive color. It is therefore desirable to characterize colors in HSI space for color measurement and color matching.
HSI is modeled with cylindrical coordinates. One possible model comprises the double cone model, i.e., two cones placed end to end or an inverted cone below another cone (see FIG.
4
). For information on the double cone model, please see “A Simplified Approach to Image Processing”, Randy Crane, Prentice Hall, 1997. The hue is represented as the angle theta, varying from 0 degree to 360 degree. Saturation corresponds to the radius or radial distance, varying from 0 to 1. Intensity varies along the z-axis with 0 being black and 1 being white. When S=0, the color is gray scale with intensity I and H is undefined. When S=1, the color is on the boundary of the top cone base and is fully saturated. When I=0, the color is black and therefore H is undefined.
On the assumption that the R, G and B values have been normalized to range from 0 to 1, the following equations may be used to convert from RGB color space to HSI (or HSL) color space:
I
=(
R+G+B
)/3
H
=
cos
-
1
{
1
2
[
(
R
-
G
)
+
(
R
-
B
)
]
[
(
R
-
G
)
2
+
(
R
-
B
)
(
G
-
B
)
]
1
2
}
S
=
1
-
3
(
R
+
G
+
B
)
[
min
(
R
,
G
,
B
)
]
The Intensity I (or Luminance L) may also be represented by the equation:
L
=0.299
R
+0.587
G
+0.114
B
which is a weighted sum of the RGB values.
The equation for H yields values in the interval [0 °, 180°]. If B/I>G/I then H is greater than 180° and is obtained as H=360°−H.
Prior art in color machine vision systems use various techniques to measure and match colors. Those skilled in the art will be familiar with ‘thresholding’ an image. To threshold a color image, a threshold is applied to each of the three planes that make up the image. In RGB mode, to select a particular color, one will need to know the red, green and blue values that make up the color. In RGB mode it is not possible to separate color from intensity. Therefore, a characterization algorithm such as histogram Intersection based on RGB space will be intensity sensitive. For more information on this, please see “Color Indexing”, Michael J. Swain, Internal Journal of Computer Vision, vol. 7:1, page 11-32, 1991.
In the HSI color space, since the color and intensity information can be separated, one usually thresholds the color image in the hue plane to identify the dominant color (hue). However, it is difficult to distinguish multiple color objects by the thresholding technique, especially when the saturation has to be taken into account. Moreover, the black and white colors are the background colors in many machine vision applications and chromaticity(i.e. hue and saturation) can not be used to represent them. Therefore, The intensity value will also have to be used to represent black and white colors in the machine vision applications.
Prior art color matching techniques commonly calculate the color difference between corresponding pixels of a target object and a template object. These prior art techniques perform pixel by pixel comparisons or subtractions between pixels of the target object and pixels of the template object. The results of these pixel by pixel comparisons may then be compiled to determine the level of color similarity between the entire target object and template object. The computation
Lin Siming
Nair Dinesh
Do Anh Hong
Meyertons Hood Kivlin Kowert & Goetzel P.C.
National Instruments Corporation
LandOfFree
System and method for color characterization with... 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 and method for color characterization with..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and System and method for color characterization with... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3299978