一种全局最优的非匀质图像分割算法

Non-homogenous image segmentation with global optimization

为了解决传统几何活动轮廓模型不能自适应地分割非匀质图像的问题,提出了一种全局优化的非匀质图像分割算法.首先,利用图像经过高斯滤波器滤波后的梯度信息定义了一个新的图像分割能量函数.然后,利用水平集方法扩展该能量函数的定义域,以使该能量函数具有全局最优解.为避免水平集函数的重新初始化过程,在能量函数中引入了一个水平集函数约束项.最后,通过最小化该能量函数,建立水平集函数演化的偏微分方程.对水平集演化方程数值求解,实现对非匀质图像的分割.实验结果表明,该算法不但能自适应地确定曲线演化方向,而且能有效地分割非匀质图像.

In order to solve the problem that the traditional geometric active contour model can not adaptively segment a non-homogenous image,a global optimization non-homogenous image segmentation algorithm is proposed.Firstly,a new energy function is defined by importing gradient information on the inhomogeneous image which is filtered by the Gaussian filter.Then,the domain of the energy function is extended by the level set method.Thus,the energy function has the solution of global optimization.We introduce a level set function control term for avoiding the re-initialization procedure of the level set function.Finally,a partial difference equation of the level set function evolvement is derived by minimizing the energy function.The non-homogenous image segmentation is implemented by the numerical solution of the partial difference equation.Experimental results show that the proposed algorithm not only can automatically determine the evolvement orientation of the active contour cure,but also can effectively segment non-homogenous images.