基于方向分数和Frangi滤波器的视网膜血管分割算法

针对视网膜图像分割中分岔和交叉等带来的难点，提出基于方向分数并结合Frangi滤波器的算法.这种算法利用基于Hessian矩阵的Frangi滤波器对线性物体的滤波效果来增强血管对比度，同时利用由蛋糕小波构造的方向分数所具有的各向异性和正交性从多角度、多方向对血管进行滤波，以利于图像的细节处理和血管网络的完整分割.该算法与其他算法的对比表明，该算法在处理分岔和交叉时优于其他算法，而且准确率、灵敏度和特异度的测量结果也都优于其他算法.

Retinal Vessel Segmentation Algorithm Based on Orientation Scores and Frangi Filter

This paper proposes an algorithm based on orientation scores combined with Frangi filter to fix the difficulties caused by bifurcation and crossings of retinal vessels. The Frangi filter， which is constructed by Hessian matrix and suitable for filtering the linear object， is used to enhance the contrast of the blood vessels.The anisotropy and orthogonality of the cake wavelet in the orientation scores are used to filter the blood vessels in multiple angles and directions， which is beneficial to the processing of details and the complete segmentation of the retinal vascular network. The comparison of the method proposed with other existing methods shows that the algorithm is better than other algorithms in processing the bifurcation and crossings， and the measurements of accuracy， sensitivity and specificity are superior to existing algorithms.