Vessel Segmentation Under Non-Uniform Illumination: A Level Set Approach

Abstract

针对血管影像中灰度不均和弱边缘情况下已有水平集模型不能正确分割血管问题,提出一种耦合了血管影像的几何信息、边缘信息和区域信息的水平集分割方法.首先,采用Hessian 矩阵的各向异性性对血管状目标进行识别,对原始影像数据进行多尺度滤波;然后采用拉普拉斯算子零交叉点的快速边缘积分方法将边缘信息嵌入能量泛函中,构建一种基于结构、边缘和区域信息的水平集分割方法.相比于单一依靠影像边缘信息或区域信息模型及其改进模型,该方法在分割严重灰度不均匀的血管造影影像上能够准确提取血管,并精确定位血管边缘.;In this paper, a new level set segementation model is proposed and is coupled with the geometric information, the edge information and the region information. The new level set segementation model is aimed at a vessel segmentation in a non-uniform image with weak object boundaries. First, a multiscaled filter with a Hessian matrix, which has a anisotropic character, is used to identify the direction of vessels. Second, the edge information is embed into a energy functional by a fast edge integral method with a laplacian zero crossing algorithm. A new level set segmentation model based on information of geometric structure, edge and region is constructed by this method. This new model can segment vessels exactly on grayscale uneven images. Compared to GAC CV segmentation model and other improved models based on CV model, the method in this paper has a better accuracy and robustness.