Post-processing for retinal vessel detection

摘要

Retinal vessel extraction plays a vital role in the computer-aided analysis of ophthalmology diseases. In this paper, we propose a new post-processing method to enhance retinal vessel classification performance. This proposed method automatically connects the discontinuous thin vessel fragments and smooth the thick vessel edges by the combination of two mathematical morphological operations, skeleton and erosion. Moreover, the proposed method removes the pathological regions by comparing the geometric structures of vessels and pathological regions. Experimental results demonstrate that the proposed method performs well for retinal vessel classification enhancement, i.e. maintain the integrity of vessel trees and reduce the false detection of pathological regions, making the vessel classification results better than those presented by the state-of-the-art approaches in comparison.