基于偏微分方程和图论两类图像分割方法的一个共同之处是将分割问题转换成了能量函数的模型建立及其最优化过程。从这一共同点出发,将图像的局部统计分布特征和Bhattacharyya相似度信息相结合并引入到测地线主动轮廓模型(GAC)和图切分(GC)模型的能量函数构造中。改进后GAC算法相当于为模型引入了一个基于似然比检验的回拉力,可有效阻止弱边界处泄露;基于非参数估计的能量函数构造更适用于小样本和分布函数不恒定的情况,使得改进GC模型更完整地提取图像目标的细节部分。将改进GAC和GC模型应用至膝关节MRI序列分割,提出完整分割各骨骼与半月板等结构的框架。在实验与分析部分,进行了定量与定性的实验对比。对噪声与局部体效应影响下的膝关节MRI序列及其他医学图像的实验,结果表明本文方法能够有效提高分割精度。

A common point of partial differential equation and graph theory based image segmentation methods lies in creating and optimizing their energy functions. From the viewpoint of creating energy models, statistical image features from nonparametric estimation are measured with Bhattacharyya metrics, which is further embedded into energy function construction in Geodesic Active Contour (GAC)and Graph Cuts (GC)models in this paper. The improved GAC and GC models benefit from the energy function based on the aforementioned metric, which introduces a pull-back strength into the GAC to prevent boundary leaking and to help the GC model in accurately estimating the distribution from small samples and unstable distribution function as well as extracting objects in more detail. Then, the proposed methods are applied to the medical image segmentation scenario and a bone and meniscus segmentation framework on knee MRI sequence is presented. In the experimental section, quantitative and qualitative comparisons are conducted respectively. Experimental results show the increased precision of our method in segmenting medical images such as knee MRI sequences, which are affected by the noise and the partial volume effect.