基于区域生长和水平集的肝脏提取分割算法

Liver Extracting and Segmenting Algorithm Based on Region Growing and Level Set

为了克服传统的以单幅图像作为信息来源的水平集模型分割复杂背景图像的局限性,结合区域生长法和水平集方法各自的特点,提出了一种新的由多幅图像信息构建的水平集分割算法模型。在运用水平集方法分割人体腹腔图像前,首先运用本文提出的一种有效的区域生长法在腹腔图像中得到肝脏的粗略分割结果作为先验形状图像。通过先验形状图像在Chan-Vese模型下控制水平集的演化,使活动轮廓的先验形状信息融合到水平集分割算法模型中,同时,利用Li模型在人体腹腔图像中进一步获取肝脏的边缘信息。这种融合多幅图像信息的复合水平集分割算法模型能够充分利用图像信息,有效地描述水平集方法中活动轮廓与目标区域肝脏的关系。通过实验验证,提出的算法模型能够很好地从人体腹腔图像中提取出肝脏区域。

To conquer the limitation of traditional level set model whose information source is just from single image when segments complex background image, a novel level set model constructed with information of multiple images by combining features of region growing method and level set method is proposed. Before segmenting the hu- man enterocoelia image, an efficient region growing method which is proposed in this paper is used to gain a rough segmentation result of liver as the prior shape image. By the prior shape image under the Chan-Vese model to con- trol livel set evolution, the active contour in the prior shape information is merged into the level set model of seg- mentation algorithm. Mean while, Li model is used further to make the edge information. The multiple level set segmentation model is merged with multiple images can fully use the image＇ s information, and effectively describe the relation of active contour and the liver which is the interest region in the level set method. By the experiment, the model which is proposed in this paper is effective to segment human enterocoelia image, and achieves the aim of extracting the liver.