一种基于iPiano非凸优化的图像分割方法

摘要

文中提出一种新的基于iPiano非凸优化算法的图像分割方法.该方法通过核函数将低维空间数据映射到高维特征空间.图像分割模型中的数据项涉及非光滑的L1范数形式,并且以Ginzburg-Landau泛函作为正则项.因此该分割模型是非凸非光滑的,采用iPiano非凸优化算法对其进行直接求解.该分割模型能够适应于多种不同的图像数据类型,不受初始条件的影响,能够获得同其他主流图像分割算法相类似的分割结果.同时该方法是iPiano非凸优化算法在图像分割领域的第一次应用,扩展了iPiano算法的应用领域.%A novel image segmentation method based on iPiano non-convex optimization algorithm is proposed.The method maps the data in lower dimensional space to a higher dimensional feature space by using a kernel function.The data term in image segmentation model involves non-smooth L1-norm form and Ginzburg-Landau functional is used as regularization term.Thus the model is non-smooth and nonconvex.The iPiano non-convex optimization algorithm is used to solve the model directly.The segmentation model can be adapt to a variety of image types and is not influenced by the initial conditions.It can get the same results as the other mainstream algorithms.Meanwhile,the method is the first application of iPiano non-convex algorithm in image segmentation.It extends the application domain of iPiano algorithm.