无需重新初始化水平集的多相水平集图像分割方法

笔者提出了一种无需重新初始化水平集函数的多相变分水平集图像分割方法.根据Vese -Chan提出的多相模型的区域竞争思想,结合图像区域全局信息的能量函数作为模型的外部能量项,引入内部变形能量约束水平集函数来逼近符号距离函数,避免了重新初始化水平集函数的过程,提高了计算速度和分割效果.实验结果表明,笔者提出的模型有效地克服了传统模型需重新初始化和构造符号距离函数计算量大的缺点,而且具有全局最优分割能力和较强的抗噪性能.

MULTIPHASE LEVEL SET IMAGE SEGMENTATION METHOD WITHOUT RE- INITIALIZATION

In this paper, we present an improved multiphase level set image segmentation method, which consists of an internal energy term forces the level set function to be close a signed distance function, and an external energy term that integrate the image information from the region. Experimental results show that the proposed model efficiently avoids the re- initialization and overcomes the problem of the computational complication, and therefore speeds up the curve evolution. Moreover, our model is able to acquire the global optimization of the segmentation and has a good anti -noise performance.