System and method for incorporating segmentation boundaries into the calculation of fractal dimension features for texture discrimination

• US 5,671,294 A
• Filed: 09/15/1994
• Issued: 09/23/1997
• Est. Priority Date: 09/15/1994
• Status: Expired due to Fees

• Current Assignee: The United States of America As Represented By The Secretary of Agriculture
• Original Assignee: the united states of america as represented by the secretary of the navy
• Inventors: Lorey, Richard A., Rogers, George W., Julin, Erik G., Priebe, Carey E., Solka, Jeffrey L.
• Primary Examiner(s): Mancuso, Joseph
• Assistant Examiner(s): Johnson, Timothy M.
• Application Number:
  US08/308,112
• Time in Patent Office: 1,104 Days
• Field of Search: 382/108, 382/131, 382/132, 382/199, 382/224, 382/228, 382/249, 382/173, 395/130
• US Class Current: 382/228
• CPC Class Codes: G06T 2207/20064 Wavelet transform [DWT]
• CPC Class Codes: G06T 2207/20076 Probabilistic image processing
• CPC Class Codes: G06T 2207/30068 Mammography; Breast
• CPC Class Codes: G06T 7/11 Region-based segmentation
• CPC Class Codes: G06T 7/143 involving probabilistic approaches, e.g. Markov random field [MRF] modelling
• CPC Class Codes: G06T 7/48 using fractals
• CPC Class Codes: G06V 10/478 Contour-based spectral representations or scale-space representations, e.g. by Fourier analysis, wavelet analysis or curvature scale-space [CSS]
• CPC Class Codes: G06V 10/52 Scale-space analysis, e.g. wavelet analysis multi-scale boundary representations G06V10/42

Abstract

Image analysis is performed by defining segmentation boundaries within an image by using wavelet theory or some other suitable method. Such boundaries can be incomplete, irregular, and/or multi-valued. The segmentation boundaries are then incorporated into feature calculations related to fractal dimensions for each pixel using a diffusion related method or a Dijkstra potential related method. These features are then used in statistical techniques to distinguish among textures or classes of interest. The system performing the image analysis is trained (or supervised) on data from different classes within an image or images. This enables the system to then later identify these classes in different images. The system can be used for Computer Aided Diagnosis (CAD) of mammograms or other medical imagery.

Claims

1. A method for distinguishing and classifying textures of an image in a supervised feature generation system, said method comprising the steps of:

generating a digitized image representative of said image;

reading and storing a digitized image file in dependence upon said digitized image;

reading and storing texture classification information representative of selected pixels within said image;

reading and storing a segmentation map representative of segmentation boundaries within said image;

recursively computing bounding areas within said image as a function of scale in dependence upon said digitized image file and said segmentation map;

computing intensity differences among bounding areas for each scale;

determining local area averages for each said scale in dependence upon said intensity differences and said segmentation map;

computing power law features;

merging said power law features and said texture classification information to produce a file of distinct and labeled features;

building a probability density function of said image using adaptive mixtures, by estimating a probability density function for each class of texture within said image and then determining class discriminant boundaries in dependence upon a maximum likelihood ratio; and

,

determining texture classes within a new digital image by comparing the probability density function of each pixel within said new digital image to previously determined discriminant boundaries and then labeling said texture classes within said new digital image with a probabilistic ranking based on probability density functions previously determined from known texture classes.

2. A method for distinguishing and classifying textures of an image in a supervised feature generation system, said method comprising the steps of:

generating a digitized image representative of said image;

reading and storing a digitized image file in dependence upon said digitized image;

reading and storing texture classification information representative of selected pixels within said image;

reading and storing a segmentation map representative of segmentation boundaries within said image;

recursively computing bounding areas within said image as a function of scale in dependence upon said digitized image file and said segmentation map;

determining weighted area averages for each said scale in dependence upon an adaptive kernel calculated from a Dijkstra potential and said segmentation map;

computing power law features for pixels within said image in dependence upon said weighted area averages;

distinguishing among textures present within said image in dependence upon said power law features; and

merging said power law features and said texture classification information to produce a file of distinct and labeled features.

3. The method for distinguishing and classifying textures as claimed in claim 2, further comprised of utilizing one of a digital scanner and a sensor to generate said digitized image.

4. The method for distinguishing and classifying textures as claimed in claim 2, further comprised of preprocessing said digitized image using contrast enhancement techniques.

5. The method for distinguishing and classifying textures as claimed in claim 2, further comprised of recursively computing said bounding areas within said image using erosion and dilation operations.

6. The method for distinguishing and classifying textures as claimed in claim 2, further comprised of defining said segmentation map using wavelet theory.

7. The method for distinguishing and classifying textures as claimed in claim 2, further comprised of projecting multiple dimensions associated with each pixel within said image to one dimension using a Fisher linear discriminant.

8. The method for distinguishing and classifying textures as claimed in claim 2, further comprised of building a probability density function model of said image using adaptive mixtures.

9. The method for distinguishing and classifying textures as claimed in claim 8, further comprised of building said probability density function model by estimating a probability density function for each class of texture within said image and then determining class discriminant boundaries in dependence upon a maximum likelihood ratio.

10. The method for distinguishing and classifying textures as claimed in claim 9, further comprised of determining texture classes within a new digital image by comparing the probability density function of each pixel within said new digital image to previously determined discriminant boundaries and then labeling said texture classes within said new digital image with a probabilistic ranking based on probability density functions previously determined from known texture classes.

11. A method for distinguishing and classifying textures of an image in an unsupervised feature generation system, said method comprising the steps of:

generating a digitized image representative of said image;

reading and storing a digitized image file in dependence upon said digitized image;

reading and storing a segmentation map representative of said image;

recursively computing bounding areas within said image as a function of scale in dependence upon said digitized image file and said segmentation map;

determining weighted area averages for each said scale in dependence upon an adaptive kernel calculated from a Dijkstra potential and said segmentation map;

computing power law features for pixels within said image in dependence upon said weighted area averages;

distinguishing among textures present within said image in dependence upon said power law features; and

generating a file of distinct unlabeled features in dependence upon said power law features.

12. The method for distinguishing and classifying textures as claimed in claim 11, further comprised of utilizing one of a digital scanner and a sensor to generate said digitized image.

13. The method for distinguishing and classifying textures as claimed in claim 11, further comprised of preprocessing said digitized image using contrast enhancement techniques.

14. The method for distinguishing and classifying textures as claimed in claim 11, further comprised of recursively computing said bounding areas within said image using erosion and dilation operations.

15. The method for distinguishing and classifying textures as claimed in claim 11, further comprised of defining said segmentation map using wavelet theory.