System and method for edge detection of an image

• US 7,502,525 B2
• Filed: 01/27/2003
• Issued: 03/10/2009
• Est. Priority Date: 01/27/2003
• Status: Active Grant

• Current Assignee: Boston Scientific Scimed, Inc. (Boston Scientific Corporation)
• Original Assignee: Boston Scientific Scimed, Inc. (Boston Scientific Corporation)
• Inventors: Sathyanarayana, Shashidhar
• Primary Examiner(s): Bella; Matthew C.
• Assistant Examiner(s): LIEW, ALEX KOK SOON
• Application Number:
  US10/352,384
• Publication Number:
  US 20040146201A1
• Time in Patent Office: 2,234 Days
• Field of Search: 382/200
• US Class Current: 382/266
• CPC Class Codes: G06T 2207/10072 Tomographic images
• CPC Class Codes: G06T 2207/10132 Ultrasound image
• CPC Class Codes: G06T 2207/30004 Biomedical image processing
• CPC Class Codes: G06T 7/13 Edge detection
• CPC Class Codes: G06T 7/194 involving foreground-background segmentation
• CPC Class Codes: G06V 10/44 Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components

Abstract

The improved system and method detects edges in an image defined by a data set. In one aspect, a plurality of arrays of pattern recognition classifiers are populated throughout an image. Each array may yield a value which indicates whether or not the array is on or near an edge to be detected. The edge may be interpolated between the arrays that are on or near an edge to be detected or extrapolated after an array on or near the edge. This edge detection method may result in more accurate edge detection.

Claims

1. A method for detecting an edge within an image defined by a data set comprising a plurality of pixels, comprising:

identifying a plurality of distinct classes based on textural differences between the plurality of distinct classes, wherein an interface on the image is defined between the plurality of distinct classes, and further wherein the edge on the image is defined by the interface;

selecting a plurality of regions of the image, each region comprising a plurality of the pixels of the image;

for each of the selected regions, applying an array of pattern recognition classifiers to the region using a processor, wherein each array comprises a plurality of pattern recognition classifiers and each pattern recognition classifier is configured and arranged to classify at least a portion of the region according to at least one of the plurality of distinct classes;

identifying each region that exhibits a predetermined response pattern in the array of pattern recognition classifiers for that region indicating an increased likelihood that the region includes at least a portion of the edge, wherein the predetermined response pattern for each region comprises a predetermined pattern of classifications by the classifiers in the array for that region; and

indicating the edge between the regions that exhibit the predetermined response pattern.

2. The method of claim 1, wherein the step of indicating the edge includes interpolating between the regions that exhibit the predetermined response pattern.

3. The method of claim 1, wherein the step of indicating the edge includes extrapolating at least a portion of the edge from at least one of the regions that exhibits the predetermined response pattern.

4. The method of claim 1, wherein each pattern recognition classifier yields a feature vector.

5. The method of claim 2, wherein interpolating the regions comprises drawing a smooth curve through the centers of the regions that exhibit the predetermined response pattern.

6. The method of claim 1, wherein applying the array of pattern recognition classifiers includes using a rectangular array of pattern recognition classifiers.

7. The method of claim 6, wherein each rectangular array comprises at least two pattern recognition classifiers.

8. The method of claim 6, wherein each rectangular array comprises at least two rows and two columns of pattern recognition classifiers.

9. The method of claim 1, wherein applying the array of pattern recognition classifiers includes using a circular array of pattern recognition classifiers.

10. The method of claim 1, wherein identifying each region includes identifying regions whose predetermined response pattern indicate that the number of pattern recognition classifiers lying on one side of the edge is approximately equal to the number of pattern recognition classifiers lying on the other side of the edge.

11. The method of claim 1, wherein identifying each region includes identifying regions whose predetermined response pattern indicate that the number of pattern recognition classifiers lying on or substantially close to the edge exceeds a threshold.

12. The method of claim 1, wherein identifying each region includes identifying regions in which a count of the number of pattern recognition classifiers lying on or substantially close to the edge exceeds a threshold.

13. The method of claim 8, wherein each rectangular array comprises at least four pattern recognition classifiers.

14. The method of claim 13, wherein each rectangular array comprises at least eight pattern recognition classifiers.

15. The method of claim 1, wherein each pattern recognition classifier applied to a region of the image returns a value identifying a class, from the plurality of distinct classes, of at least a portion of the region.

16. The method of claim 1, wherein each pattern recognition classifier utilizes a Bayesian classification technique for classification of a region.