摘要
针对条件随机场(CRF)模型在参数估计及模型推断阶段时间复杂度较高的问题,引入简单线性迭代聚类(SLIC)的超像素方法,提出一种基于SLIC的条件随机场图像分割算法。该算法首先通过SLIC对图像进行预处理,将图像划分成内部相似性较高的超像素区域,然后以超像素作为节点建立CRF图模型,最后通过参数估计及模型推断获得图像分割结果。实验结果表明,基于SLIC的条件随机场图像分割模型在获得较好分割结果的同时,极大缩短了运行时间,提高了分割的效率。
The stage of estimating parameter and inferencing model in conditional random field （CRF） has a high time complexity. To improve this problem, this paper introduced the conception of simple linear interactive clustering （ SLIC ） method into the CRF model, and proposed a SLIC-based CRF image segmentation algorithm. Firstly, the algorithm pre-segmented the image into small homogeneous super-pixel regions by using SLIC method. Then it constructed the CRF graphical model with super-pixels as nodes. Finally, it obtained the segmentation results by estimating parameter and inferencing model. The experiments show that the proposed algorithm can obtain good results. At the same time, it can reduce the running time largely and improve the efficiency.
出处
《计算机应用研究》
CSCD
北大核心
2015年第12期3817-3820,3824,共5页
Application Research of Computers
基金
国家自然科学基金资助项目(10974130)
中央高校基本科研业务费专项资金资助项目(GK201405007)
关键词
条件随机场
简单线性迭代聚类
超像素
图像分割
参数估计
conditional random field
simple linear interactive clustering
super-pixel
image segmentation
parameter estimation
