Retinal vessel image segmentation based on correlational open active contours model

As the local contrast of retinal blood vessel is unstable, especially in unhealthy fundus images, an open active contour based method is presented for automated segmentation of retinal vessels. Firstly, we proposed a new initial parameters giving method and a pair of correlation open active contour model, based on these method, each vessel edge is modeled with an active contour which initiated by the corresponding boundary of Hessian vessel response; Secondly, to avoid the problem of snake from falling into the local minima, a local region pixel representation method named pixel´s average intensity influence was proposed. Finally, a post-processing procedure based on context feature is introduced to filtering the non-vessel linear structures. The effectiveness of the proposed method was demonstrated through receiver operating characteristic analysis on the benchmarked dataset of DRIVE. The method achieves an average area under the receiver operating characteristic curve of 0.9543. Compared with some the state-of-the-art, the results of this method proposed show that our method is significantly better than the others. For example, the average accuracy, sensitivity and specificity achieved on the DRIVE are 0.9521, 0.7508, 0.9656, respectively.