Image analysis – Color image processing – Image segmentation using color
Reexamination Certificate
1999-08-30
2003-03-18
Boudreau, Leo (Department: 2621)
Image analysis
Color image processing
Image segmentation using color
C382S162000, C382S165000, C382S167000, C382S173000, C382S205000, C382S254000, C382S260000, C382S266000, C382S282000, C358S515000, C358S520000, C358S538000
Reexamination Certificate
active
06535632
ABSTRACT:
BACKGROUND OF THE INVENTION
This invention relates to color image processing techniques such as object tracking and image segmentation, and more particularly to a process for filtering HSI data for object tracking and image segmentation.
Color image processing techniques often are used in image enhancement, video encoding, video editing and computer vision applications. Image tracking relates to the identification of an image object each frame in a sequence of image frames, such as in a sequence of motion video frames. Image segmentation is used to identify boundaries and edges of image objects in an image frame.
HSI refers to the Hue, Saturation, Intensity color model for presenting color data. There are many different color models (also referred to as color domains or color spaces) developed for the representation and manipulation of color data. Color monitors typically use a Red, Green, Blue (RGB) color model. Color printers typically use a Cyan, Yellow, Magenta (CYM) or a Cyan, Yellow, Magenta, Black (CYMK) color model. Color television broadcast signals typically use a luminance, intensity, color difference (YIQ) color model, where I and Q relate to chrominance.
The Hue Saturation Intensity (HSI) color model closely resembles the color sensing properties of human vision. The intensity component is related to the luminance component decoupled from the color. The hue and saturation components are related to the way in which a human perceives color. Such relation to human vision makes it desirable to use the HSI color model for color image processing techniques, such as image enhancement and image segmentation.
The input image data for color image processing techniques typically is in RGB format. Unfortunately the transformation from RGB to HSI color space and from HSI to RGB color space is very nonlinear and complicated in comparison to the conversion formulas among the other color models. As an example, when an RGB image is degraded by random noise, the nonlinearity in the conversion formulae causes the noise distribution in HSI color space to be nonuniform. Further, the noise distribution in HSI color space depends on the intensity and saturation values of the input data. For example, when the intensity value is small, the noise in the saturation and hue is large. This creates problems in using the HSI color model for image processing techniques, such as image enhancement and image segmentation. Accordingly, there is a need for a method which reduces the magnitude of the noise or the nonuniformity of the noise variance in HSI color space.
With regard to object tracking, it is known to use data clustering methods, such as found in pattern learning and recognition systems based upon adaptive resonance theory (ART). Adaptive resonance theory, as coined by Grossberg, is a system for self-organizing stable pattern recognition codes in real-time data in response to arbitrary sequences of input patterns. (See “Adaptive Pattern Classification and Universal Recoding: II . . . ,” by Stephen Grossberg, Biological Cybernetics 23, pp. 187-202 (1976)). It is based on the problem of discovering, learning and recognizing invariant properties of a data set, and is somewhat analogous to the human processes of perception and cognition. The invariant properties, called recognition codes, emerge in human perception through an individual's interaction with the environment. When these recognition codes emerge spontaneously, as in human perception, the process is said to be self-organizing.
With regard to image segmentation, active contour models, also known as snakes, have been used for adjusting image features, in particular image object boundaries. In concept, active contour models involve overlaying an elastic curve onto an image. The curve (i.e., snake) deforms itself from an initial shape to adjust to the image features. An energy minimizing function is used which adapts the curve to image features such as lines and edges. The function is guided by external constraint forces and image forces. The best fit is achieved by minimizing a total energy computation of the curve. The energy computation is derived from (i) energy terms for internal tension (stretching) and stiffness (bending), and (ii) potential terms derived from image features (edges; corners). A pressure force also has been used to allow closed contours to inflate. Conventionally, iterations are applied to get the entire contour to converge to an optimal path.
SUMMARY OF THE INVENTION
According to the invention, adaptive noise filtering is applied to an image frame of HSI data to reduce and more uniformly distribute noise while preserving image feature edges. In one implementation for a sequence of image frames, such filtering allows for improved image object tracking ability and improved image object segmentation.
According to one aspect of the invention, it has been found that in transforming an RGB image into HSI color space, noise present in the RGB image is nonuniformly distributed within the resulting HSI image. In particular the hue and saturation components have what may be considered to be a Cauchy distribution of noise where mean and variance do not exist. As a result, a noise distribution model has been determined experimentally.
According to another aspect of this invention, the HSI data is filtered using an adaptive spatial filter having a plurality of averaging kernels. An appropriate kernel is selected for each pixel for each of the hue and saturation components. A set of thresholds are defined for selecting the kernel for the hue component. Another set of thresholds are defined for selecting the kernel for the saturation component.
According to another aspect of this invention, the kernel for the saturation component is selected by comparing the intensity component to the saturation component thresholds.
According to another aspect of this invention, the kernel for the hue component is selected by comparing the product of intensity component and the saturation component to the hue component thresholds.
According to another aspect of this invention, a color gradient operation is applied to the filtered HSI data to aid in detecting image object boundaries.
According to another aspect of the invention, a method is provided for segmenting an image frame of pixel data, in which the image frame includes a plurality of pixels. For each pixel of the image frame, the corresponding pixel data is converted into hue, saturation, intensity color space. The HSI pixel data then is filtered with the adaptive spatial filters. Object segmentation then is performed to define a set of filtered HSI pixel data corresponding to the image object. The image frame then is encoded in which pixel data corresponding to the image object is encoded at a higher bit rate than other pixel data.
An advantage of the invention is that image segmentation techniques are performed in HSI color space where color sensing properties more closely resemble human vision. According to another advantage of this invention, object boundaries are preserved while noise level is significantly reduced and the noise variance is made more uniform.
REFERENCES:
patent: 4819077 (1989-04-01), Kikuchi et al.
patent: 5949914 (1999-09-01), Yuen
patent: 5959693 (1999-09-01), Wu et al.
patent: 5995164 (1999-11-01), Tsyrganovich
patent: 6226401 (2001-05-01), Yamafuji et al.
patent: 6229580 (2001-05-01), Inoue
Mital, et al discloses “colour vision for industrial applications”, IEEE, pp. 519-522, 1989.*
Ladley, et al discloses “fundamentals of true-color image processing”, IEEE, pp. 791-795, 1990.*
Lin, et al discloses “color image segmentation using modified HSI system for road following”, IEEE, pp. 1998-2003, 1991.*
Welch, et al discloses “image processing using the HSI color space”, IEEE, pp. 722-725, 1991.*
Parsiani discloses “isolated noise detection and reduction in color images”, IEEE, pp. 741-743, 1993.*
Sobottka, et al discloses “segmentation and tracking of faces in color images”, IEEE, pp. 236-241, 1996.*
Sparling, et al discloses “photoelastic analysis using ch
Gopishankar Lakshmanan
Kim Yongmin
Park Hyun Wook
Sun Shijun
Boudreau Leo
Koda Esq. Steven P.
Mariam Daniel G.
University of Washington
LandOfFree
Image processing in HSI color space using adaptive noise... 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 in HSI color space using adaptive noise..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Image processing in HSI color space using adaptive noise... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3085692