Fig 1.
Step-wise illustration of the proposed system.
Fig 2.
Histogram based visual representation of the applied preprocessing contrast enhancement schemes.
(a) Green channel image (b) CLAHE (c) TopHat image (d) Frangi enhanced image.
Fig 3.
Analysis of Frangi filtering enhancement using DRIVE dataset.
(a) Thin vessel enhanced image (b) Thin binary image (c) Thick vessel enhanced image (d) Thick binary image.
Fig 4.
Analysis of Frangi filtering enhancement using STARE dataset.
A (a) Thin vessel enhanced image (b) Thin binary image (c) Thick vessel enhanced image (d) Thick binary image.
Fig 5.
2D histogram top view.
Fig 6.
Steps for extraction of VLM.
Fig 7.
Analysis of using OR operator.
(a) RGB input image (b) VLM (c) Frangi enhanced, thresholded image (d) OR Operator result.
Fig 8.
Visual presentation of the proposed system major processing stages.
(a) Input RGB photograph from STARE database (b) Green channel (c) CLAHE applied result (d) Difference image (e) Otsu threshold resultant image (f) Postprocessed dilated image (g) Frangi filter enhanced image (h) Final image using AND Operation.
Fig 9.
Visual presentation of the proposed system major processing stages.
(a) Input RGB photograph from DRIVE database (b) Green channel (c) CLAHE applied result (d) Difference image (e) Otsu threshold resultant image (f) Postprocessed dilated image (g) Frangi filter enhanced image (h) Final image using AND Operation.
Fig 10.
Visual presentation of the Proposed system major processing stages.
(a) Input RGB photograph from HRF database (b) Green channel (c) CLAHE applied result (d) Difference image (e) Otsu threshold resultant image (f) Postprocessed dilated image (g) Frangi filter enhanced image (h) Final image using AND Operation.
Table 1.
Performance metrics for evaluation of the proposed method.
Table 2.
Accuracy (Acc), Sensitivity (Sn) and Specificity (Sp) statistics of the proposed system on the DRIVE, STARE and HRF databases.
Table 3.
Performance evaluations of various retinal vascular extraction algorithms.
Fig 11.
Visual appearance of the proposed technique utilizing STARE dataset.
(a) RGB photograph (b) Manual segmentation (c) Proposed technique segmented image.
Fig 12.
Visual appearance of the proposed technique utilizing DRIVE dataset.
(a) RGB photograph (b) Manual segmentation (c) Proposed technique segmented image.
Fig 13.
Pictorial representation for unhealthy retinal image from the STARE dataset.
(a) RGB image (b) Manual segmentation (c) Proposed scheme final result.
Fig 14.
Pictorial representation for unhealthy retinal image from the DRIVE dataset.
(a) RGB image (b) Manual segmentation (c) Proposed scheme final result.
Table 4.
Proposed system segmentation results assessment with different retinal extraction approaches for abnormal retinal images.
Fig 15.
Proposed system results with and without using VLM on pathological (first two rows) and normal (last two rows) images from the STARE dataset.
(a) Color image (b) Manual Segmented image (c) Proposed system final image without VLM (d) Segmentation results with utilizing VLM.
Table 5.
Time complexity assessment of various methods with the proposed method.