SEGMENTATION AND IDENTIFICATION OF CLOSED-CONTOUR FEATURES IN IMAGES USING GRAPH THEORY AND QUASI-POLAR TRANSFORM

• US 20150371400A1
• Filed: 01/27/2014
• Published: 12/24/2015
• Est. Priority Date: 01/25/2013
• Status: Active Grant

• Current Assignee: Duke University
• Original Assignee: Duke University
• Inventors: Farsiu, Sina, Chiu, Stephanie J., Izatt, Joseph A.
• Granted Patent: US 9,940,722 B2
• Time in Patent Office: Days
• US Class Current: 1/1
• CPC Class Codes: G06T 2207/10056 Microscopic image
• CPC Class Codes: G06T 2207/10101 Optical tomography; Optical coherence tomography [OCT]
• CPC Class Codes: G06T 2207/20072 Graph-based image processing
• CPC Class Codes: G06T 2207/30041 Eye; Retina; Ophthalmic
• CPC Class Codes: G06T 3/00 Geometric image transformations in the plane of the image
• CPC Class Codes: G06T 5/00 Image enhancement or restoration
• CPC Class Codes: G06T 7/0012 Biomedical image inspection
• CPC Class Codes: G06T 7/12 Edge-based segmentation
• CPC Class Codes: G06T 7/13 Edge detection
• CPC Class Codes: G06T 7/162 involving graph-based methods

Abstract

Segmentation and identification of closed-contour features in images using graph theory and quasi-polar transform are disclosed. According to an aspect, a method includes representing, in a rectangular domain, an image including a feature of interest. Further, the method includes determining a point within the feature of interest. The method also includes transforming the image of the feature from the rectangular domain to a quasi-polar domain based on the point. The quasi-polar domain is represented as a graph of nodes connected together by edges. The method also includes graph cutting the quasi-polar domain to identify the boundary of the feature of interest in the image.

Claims

1. A method comprising:

using at least one processor and memory for;

representing, in a rectangular domain, an image including a feature of interest;

determining a point within the feature of interest;

transforming the image of the feature from the rectangular domain to a quasi-polar domain based on the point, the quasi-polar domain being represented as a graph of nodes connected together by edges; and

graph cutting the quasi-polar domain to identify the boundary of the feature of interest in the image.

2. The method of claim 1, further comprising using the at least one processor and memory for estimating a boundary of the feature of interest.

3. The method of claim 2, wherein determining the point comprises determining the point within the estimated boundary of the feature of interest.

4. The method of claim 1, further comprising using the at least one processor and memory for:

adding, to the graph, nodes adjacent to nodes along first and second sides of the graph, wherein the added nodes have edge weights less than the nodes along the first and second sides of the graph; and

assigning start and end points to any of the added nodes along the first and second sides, respectively; and

wherein graph cutting comprises graph cutting between the start and end points to identify the feature of interest in the image.

5. The method of claim 4, further comprising using the at least one processor and memory for shifting the nodes along the path to be flat.

6. The method of claim 4, further using the at least one processor and memory for comprising assigning the edge weights based on an accumulate gradient of nodes corresponding to the edges.

7. The method of claim 1, further comprising using the at least one processor and memory for:

adding, to the graph, a first node and a second node adjacent to nodes along a first side and a second side, respectively, of the graph; and

wherein graph cutting comprises graph cutting between the first and second nodes to identify the feature of interest in the image.

8. The method of claim 1, wherein the feature of interest is one of a closed-contour feature and a substantially closed-contour feature.

9. The method of claim 1, wherein the feature of interest is a retinal closed-contour feature.

10. The method of claim 1, wherein graph cutting comprises applying a shortest path algorithm to determine a minimum-weighted path across the graph between the start and end nodes, wherein the minimum-weighted path corresponds to the feature of interest.

11. The method of claim 1, further comprising using the at least one processor and memory for:

filtering a digitized image of the feature of interest to generate a binary image; and

manipulating the binary image using morphological operators to reduce or eliminate noise of artifacts in the binary image; and

wherein transforming the image comprises transforming the binary image to the quasi-polar domain.

12. The method of claim 1, further comprising using the at least one processor and memory for shifting pixels on the transformed image until the feature of interest is substantially flat.

19. The computer-implemented system of claim 1, wherein the at least one processor and memory are configured to:

add, to the graph, a first node and a second node adjacent to nodes along a first side and a second side, respectively, of the graph; and

wherein graph cutting comprises graph cutting between the first and second nodes to identify the feature of interest in the image.

13. A computer-implemented system comprising:

at least one processor and memory configured to;

represent, in a rectangular domain, an image including a feature of interest;

determine a point within the feature of interest;

transform the image of the feature from the rectangular domain to a quasi-polar domain based on the point, the quasi-polar domain being represented as a graph of nodes connected together by edges; and

graph cut the quasi-polar domain to identify the boundary of the feature of interest in the image.

14. The computer-implemented system of claim 13, wherein the at least one processor and memory are configured to estimate a boundary of the feature of interest.

15. The computer-implemented system of claim 14, wherein the at least one processor and memory are configured to determine the point within the estimated boundary of the feature of interest.

16. The computer-implemented system of claim 13, wherein the at least one processor and memory are configured to:

add, to the graph, nodes adjacent to nodes along first and second sides of the graph, wherein the added nodes have edge weights less than the nodes along the first and second sides of the graph;

assign start and end points to any of the added nodes along the first and second sides, respectively; and

wherein graph cutting comprises graph cutting between the start and end points to identify the feature of interest in the image.

17. The computer-implemented system of claim 16, wherein the at least one processor and memory are configured to shift the nodes along the path to be flat.

18. The computer-implemented system of claim 16, wherein the at least one processor and memory are configured to assign the edge weights based on an accumulate gradient of nodes corresponding to the edges.

20. The computer-implemented system of claim 13, wherein the feature of interest is one of a closed-contour feature and a substantially closed-contour feature.

21. The computer-implemented system of claim 13, wherein the feature of interest is a retinal closed-contour feature.

22. The computer-implemented system of claim 13, wherein graph cutting comprises applying a shortest path algorithm to determine a minimum-weighted path across the graph between the start and end nodes, wherein the minimum-weighted path corresponds to the feature of interest.

23. The computer-implemented system of claim 13, wherein the at least one processor and memory are configured to:

filter a digitized image of the feature of interest to generate a binary image; and

manipulate the binary image using morphological operators to reduce or eliminate noise of artifacts in the binary image; and

wherein transforming the image comprises transforming the binary image to the quasi-polar domain.

24. The computer-implemented system of claim 13, wherein the at least one processor and memory are configured to shift pixels on the transformed image until the feature of interest is substantially flat.