基于全局主动轮廓模型的图像分割研究

ON GLOBAL ACTIVE CONTOUR MODEL-BASED IMAGE SEGMENTATION

基于区域信息的Mumford-shah分割模型中,由于图像对象灰度分布的不一致性,分割曲线易陷入局部最小值,不能获得完整边界分割。提出了一个全新的主动轮廓模型,基于局部区域信息创建能量驱动曲线演化,通过竞争关系约束曲线的演化范围,确保模型能收敛于全局静态最小值。实验结果表明,分割模型可以同时分割多个灰度分布不均匀的对象,对噪声具有较好的鲁棒性。

In Mumford-Shah segmentation model based on region information, because of the inhomogeneity of grey distribution of the image object, the segmenting curve often gets into local minimum and can＇ t segment intact boundary. In this paper we propose a novel active contour model, in which the evolutional curve is driven by the energy created with local statistical information, and the evolution range is re- strained by the competition between the curves, so that the curve is ensured to converge to a global stationary minimum. Experiment results show that our novel active contour model can segment muhi-objects with inhomogeneous grey distribution simultaneously, and is robust to noisy images.