Abstract: This paper proposes and establishes a multi-scale statistic model scheme based on wavelet analysis, and adds it to the Mumford-Shah energy function for the better of the segmentation based shape prior. The experiments show the result is significant in special use of medical images with complex topological structure, the accuracy is also improved as well as it’s computing velocity

Key words: Level set; Image segmentation; Wavelet; Mumford-Shah funcition; Shape prior; Statistic model

摘要： 提出并建立了一种基于小波分析的多尺度统计模型，将该统计模型作为先验知识引入Mumford-Shah 能量约束函数，从而指导水平集函数进行图像分割。实验表明，当对拓扑结构复杂的医学图像进行分割时，该方法具有明显的效果，同时分割速度和精度都得到了明显改善

关键词: 水平集；图像分割；小波；Mumford-Shah 能量函数；先验知识；统计模型