An Algorithm for Segmentation of Blood Vessels in Images

Abstract

Segmentation is a process of partitioning an image into separate regions until the ones of interest are singled out. We present an algorithm that combines the approaches of thresholding and region growing to segment the vessel tree. First, the optimal global threshold is determined experimentally. We found that the ordinary mean of all the intensity values in the image yields the best estimate for the value of the global threshold. The mean correlated with the threshold value obtained experimentally better than the other two statistics investigated: the sum and the standard deviation of all the intensity values. Upon derivation of the global threshold, it is applied to the original image to yield the seed points. Those are the initial points for the region growing procedure. Then, the algorithm determines the local threshold that serves as the stop condition for the region-growing procedure. Next, the value for the local threshold is obtained from the histogram of the gray-level values. This is equivalent to finding the point of the maximum curvature on the right slope of the histogram‘s second hill. Ill. 9, bibl. 5 (in English; summaries in English, Russian and Lithuanian).