Hessian-based vessel enhancement combined with directional filter banks and vessel similarity

Abstract

Vessel enhancement in angiogram is an important preprocessing procedure for clinical diagnosis and further processing. Hessian-based methods have been widely used due to its elegant geometrical interpretations. But classical Hessian-based methods suffer the drawbacks of blurred vessel edges and intensity inhomogeneity, especially in vessel junction regions. In this paper, a new vessel enhancement method is proposed to address these problems. The input angiogram is first decomposed into several directional images and each of them is enhanced by traditional Hessian-based method. Then the enhanced direction images are recombined to generate final result. To improve contrast at vessel edges and intensity homogeneity in vessels, a vessel similarity measurement is proposed and used as weight coefficients in the recombination of enhanced directional images. It is calculated based on the intensity and curvature analysis of directional images. Qualitative and quantitative experiments performed on typical angiograms and retinal images in DRIVE database show that the proposed method has better performance than other two Hessian-based enhancement methods.