结合图像局部信息和Hausdorff距离的活动轮廓模型医学图像分割算法

Active Contour Method based on Region-Scalable Fitting and Hausdorff Distance for Medical Image Segmentation

本文提出了一种新的基于距离局部信息的活动轮廓摸型。该模型的能量函数将区域可扩展能量项(region scalable fitting,RSF)和Hausdorff距离项结合,其中RSF项在目标边缘附近起主导作用,用来吸引水平集函数曲线到达目标边界;而Hausdorff距离由于包含了局部区域的相似信息,可以提高分割方法的稳定性。在保证分割精度的情况下,相对于区域可伸缩拟合及局部巴氏距离的活动轮廓模型RSFB,本文模型具有更快的收敛速度和更好的参数选择鲁棒性,对于解决图像分割中的边界模糊和噪声问题效果显著。实验结果显示本文提出的模型在超声图像和不均匀图像的分割中都有非常好的效果,且计算量较小。

In this study,a new local distance-based active contour model is proposed. An energy function based on the regional-scalable fitting term and local hausdorff distance term is formulated. The RSF term is dominant near object boundary and responsible for attracting the level set contour towards object boundaries,and local hausdorff distance term including the local region similarity information improves the robustness of the proposed method. Compared to the active contour method driven by region-scalable fitting and local Bhattacharyya distance energies（ RSFB）,the proposed method has the relatively fast convergence and improves the robustness to the parameter selection. Our method can overcome the weak boundaries and noise problem in images. A level set function is used to define the partition of image domain into two disjoined regions. Experiment results demonstrate the desirable performance of the proposed method with relativity less iterations and time.