参数化水平集活动轮廓模型的快速图像分割算法

摘要 为了提高图像分割的速度，提出一种参数化水平集活动轮廓模型的快速图像分割算法.该算法中的水平集函数由参数向量确定，而非带符号距离函数，降低了水平集函数的维度.将参数化的水平集函数嵌入到经典的LGDF(local Gaussian distribution fitting)模型中进行图像分割，不需要重新初始化和额外的正则项，同时可选择较大迭代步长.实验结果表明:所提方法能够有效地分割超声、CT和核磁等医学图像，与带有正则项的分割算法LGDF和最近提出的快速分割算法MSLCV相比，在保证分割精度的同时，计算速度得到了明显提高.
Abstract：In order to improve the segmentation speed， a fast image segmentation method based on parametric level set active contour model was proposed. The level set function was determined by the parameter vector， rather than the signed distance function， which reduces the dimension of the level set function. The parametric level set function was embedded into the classical LGDF(local Gaussian distribution fitting) segmentation algorithm， and it does not need to be re-initialized or additional regular terms， and it can choose larger step length. The experiment results show that the proposed method can effectively segment medical images such as ultrasound， CT and MR medical images. Compared with the LGDF model with regular terms and the recently proposed fast segmentation algorithm MSLCV， in the case of similar segmentation accuracy， the calculation speed of the proposed method is improved obviously.