System and method for robust optic disk detection in retinal images using vessel structure and radon transform

1. A method for optic disk detection in retinal images, comprising:

extracting a vessel tree from a retinal image;

locating an optic disk in the retinal image using the vessel tree;

enhancing a contrast of the optic disk;

removing vessels from the contrast enhanced optic disk;

detecting a boundary of the optic disk with vessels removed by generating an edge map;

projecting the edge map at a plurality of angles by using a radon transform; and

estimating a radius and center of the optic disk using the projections,

wherein locating an optic disk in the retinal image comprises:

selecting first start and end points and second start and end points from different branches of the vessel tree;

finding a shortest path between the first start and end points and a shortest path between the second start and end points;

finding an intersecting point of the shortest paths at a root of the vessel tree; and

defining the intersecting point as a location of the optical disk.

2. The method of claim 1 , wherein extracting a vessel tree from a retinal image comprises:

applying shade correction and contrast enhancement to the retinal image; and

performing a binarization on the shade corrected and contrast enhanced retinal image.

3. The method of claim 1 , wherein selecting first start and end points and second start and end points from different branches of the vessel tree comprises;

defining a top-left point in the vessel tree as the first start point;

defining a bottom-right point in the vessel tree as the first end point;

defining a top-right point in the vessel tree as the second start point; and

defining a bottom-left point in the vessel tree as the second end point.

4. The method of claim 1 , wherein the shortest paths are found by using a searching algorithm.

5. The method of claim 1 , further comprising:

cropping the optic disk before enhancing a contrast of the optic disk.

6. The method of claim 5 , wherein removing vessels from the contrast enhanced optic disk comprises:

applying a morphological filter to the cropped optic disk.

7. The method of claim 6 , wherein generating the edge map comprises:

applying an edge detection filter to the cropped optic disk.

8. The method of claim 7 , further comprising:

applying a threshold to the edge map to obtain a binary map.

9. The method of claim 1 , further comprising:

before estimating a radius and center of the optic disk using the projections, identifying a portion of an edge map projected at one of the angles as the boundary of the optic disk; and

trimming left and right sides of the boundary of the optic disk.

10. A method for optic disk detection in retinal images, comprising:

extracting a vessel tree from a retinal image;

locating an optic disk in the retinal image using the vessel tree;

enhancing a contrast of the optic disk;

removing vessels from the contrast enhanced optic disk;

detecting a boundary of the optic disk with vessels removed by generating an edge map;

projecting the edge map at a plurality of angles by using a radon transform; and

estimating a radius and center of the optic disk using the projections,

wherein before estimating a radius and center of the optic disk using the projections, identifying a portion of an edge map projected at one of the angles as the boundary of the optic disk; and

trimming left and right sides of the boundary of the optic disk, and

wherein when estimating a radius and center of the optic disk using the projections,

computing a mean and standard deviation of each of three sequences, each sequence including an estimated value of a center or radius of the optic disk using the Radon transforms projected at each of the angles;

removing values from each of the sequences that are not with two standard deviations from each sequence's mean; and

estimating the radius and center of the optic disk by taking the mean of the remaining values of these sequences.

11. A system for optic disk detection in retinal images, comprising:

a memory device for storing a program;

a processor in communication with the memory device, the processor operative with the program to:

extract a vessel tree from a retinal image;

locate an optic disk in the retinal image using the vessel tree;

enhance a contrast of the optic disk;

remove vessels from the contrast enhanced optic disk;

detect a boundary of the optic disk with vessels removed by generating an edge map;

project the edge map at a plurality of angles by using a radon transform; and

estimate a radius and center of the optic disk using the projections,

wherein when locating an optic disk in the retinal image the processor is further operative with the program to:

select first start and end points and second start and end points from different branches of the vessel tree;

find a shortest path between the first start and end points and a shortest path between the second start and end points;

find an intersecting point of the shortest paths at a root of the vessel tree; and

define the intersecting point as a location of the optical disk.

12. The system of claim 11 , wherein when extracting a vessel tree from a retinal image the processor is further operative with the program to:

apply shade correction and contrast enhancement to the retinal image; and

perform a binarization on the shade corrected and contrast enhanced retinal image.

13. The system of claim 11 , wherein when selecting first start and end points and second start and end points from different branches of the vessel tree the processor is further operative with the program to:

define a top-left point in the vessel tree as the first start point;

define a bottom-right point in the vessel tree as the first end point;

define a top-right point in the vessel tree as the second start point; and

define a bottom-left point in the vessel tree as the second end point.

14. The system of claim 11 , wherein the shortest paths are found by using a searching algorithm.

15. The system of claim 11 , wherein the processor is further operative with the program to:

crop the optic disk before enhancing a contrast of the optic disk.

16. The system of claim 15 , wherein when removing vessels from the contrast enhanced optic disk the processor is further operative with the program to:

apply a morphological filter to the cropped optic disk.

17. The system of claim 16 , wherein when generating the edge map the processor is further operative with the program to:

apply an edge detection filter to the cropped optic disk.

18. The system of claim 17 , wherein the processor is further operative with the program to:

apply a threshold to the edge map to obtain a binary map.

19. The system of claim 11 , wherein before estimating a radius and center of the optic disk using the projections the processor is further operative with the program to:

identify a portion of an edge map projected at one of the angles as the boundary of the optic disk; and

trim left and right sides of the boundary of the optic disk.

20. A system for optic disk detection in retinal images, comprising:

a memory device for storing a program;

a processor in communication with the memory device, the processor operative with the program to:

extract a vessel tree from a retinal image;

locate an optic disk in the retinal image using the vessel tree;

enhance a contrast of the optic disk;

remove vessels from the contrast enhanced optic disk;

detect a boundary of the optic disk with vessels removed by generating an edge map;

project the edge map at a plurality of angles by using a radon transform; and

estimate a radius and center of the optic disk using the projections,

wherein before estimating a radius and center of the optic disk using the projections the processor is further operative with the program to:

identify a portion of an edge map projected at one of the angles as the boundary of the optic disk; and

trim left and right sides of the boundary of the optic disk, and

wherein when estimating a radius and center of the optic disk using the projections, the processor is further operative with the program to:

compute a mean and standard deviation of each of three sequences, each sequence including an estimated value of a center or radius of the optic disk using the Radon transforms projected at each of the angles;

remove values from each of the sequences that are not with two standard deviations from each sequences mean; and

estimate the radius and center of the optic disk by taking the mean of the remaining values of these sequences.