Micro-blood vessel detection using k-means clustering and morphological thinning

Abstract

This paper introduces a combination method for blood vessel segmentation based on k-means clustering and morphological thinning. In the first stage, the original image was partitioned into two clusters (foreground and background). As this step is a coarse classification, a fine detection proceeded to the pre-processed image with the help of the morphological thinning algorithm. Experimental results indicated that blood vessels within an image have been detected by using the coarse-to-fine segmentation method with the accuracy of more than 90%. © 2011 Springer-Verlag.