摘要： 基于分段常数水平集方法的M-S模型能解决二值M-S模型及其相应的C-V模型不适合包含多个目标(或多相)的图像分割的问题。采用该方法研究青光眼视乳头图像杯盘分割难题，并在此基础上进行杯盘重建。实验表明，该方法能够正确分割不同青光眼病人的视乳头图像。与多层分割方法相比，该方法能同时获得视杯和视盘的形态。

关键词: 图像分割, Mumford-Shah模型(M-S模型), 分段常数水平集方法, 青光眼

Abstract: The binary Mumford-Shah(M-S) model and its corresponding C-V model can not work if there are more than one objects (or multi-phase) in the image. M-S model based on Piecewise Constant Level Set Method(PCLSM) can solve the minimal image partition problem. The PCLSM based Mumford-Shah model is applied to the segmentation of optic cup and disk in optic nerve head images. On this basis, the image of optic cup and disk is reconstructed. The tests demonstrate that this technique can segment simultaneously the optic cup and disk in the optic nerve head images from the different glaucoma patients, which is different from hierarchical M-S model.

Key words: image segmentation, Mumford-Shah model(M-S model), Piecewise Constant Level Set Method(PCLSM), glaucoma

中图分类号: 

• TP391