Morphological Scale-Space based Vessel Segmentation of Retinal Image

This paper introduces a novel procedure to segment retinal vessels using new technique namely Morphological Angular Scale-Space (MASS). Line structuring element is rotated about the seed point to determine the curvature of the vessels thereby ensuring that the components remains connected along vessels segmented. Scale-Space is created by varying the length of the structuring element which gradually reduces non-vessel like elements from the processing image. Information from the lower scale image been given as a feedback to build the image of higher scales thereby extracting and retaining vessels effectively from lower scales moving to higher scale-space. The method, attains lowest mean square error value at a certain scale, which is observed to be better than most of the other morphological techniques of vessel segmentation. A publicly available DRIVE database is used for analysis of the MASS technique as well as comparing with other methods.