SIFT detector based color retinal fundus image segmentation.

Computerized segmentation of retinal vessels can help doctors discover diabetic retinopathy, Retinal Tear and Retinal Detachment. Because the anatomy of retinal vessels is complicated and changing, automated vessel segmentation is still a difficult task. The Expectation-Maximization approach for retinal vascular segmentation is proposed in this paper, which uses a matched filter (MF) that combines Mask based Blood Vessels Extraction with a global image threshold filter to extract the vessels network. Firstly, the Contrast Limited Adaptive Histogram Equalization method is applied to the retinal image in order to improve the contrast between the blood vessels and the background. Scale Invariant Feature Transform Detector is used to estimate the vessel key point. The expectation maximization technique is used to extract retinal blood vessels, which uses a mask-based blood vessel extraction with a global picture threshold. The evaluation was assessed utilizing ANN to evaluate segmentation performance with Normal and Abnormal Images. The performance of algorithm is compared and analyzed using other approaches. [ABSTRACT FROM AUTHOR]