基于相位场模型的提花织物图像分割算法  

Segmentation algorithm for jacquard image based on phase field model

作　　者：吴洪森[1] 冯志林[2] 周佳男[2] 刘小明[3] 董金祥[3] 

机构地区：[1]浙江警察学院基础部,浙江杭州310053 [2]浙江工业大学信息工程学系,浙江杭州310024 [3]浙江大学计算机科学与工程学系,浙江杭州310037

出　　处：《浙江大学学报（工学版）》2008年第3期444-449,共6页Journal of Zhejiang University：Engineering Science

基　　金：公安部应用创新基金资助项目(2006YYCXZJST139)

摘　　要：将相变理论中的相位场模型引入到图像分割建模中,并对经典相位场模型进行改进,提出一种基于相位场模型的提花织物图像分割算法.算法采用有限元逼近和自适应网格调整技术对模型进行数值求解;采用一阶拉格朗日有限元对模型进行空间离散、半隐式欧拉格式进行时间离散.为了提高有限元网格单元对图案内容的表征能力,采用网格优化策略对网格单元进行自适应调整.采用后置误差估测子对逼近解和初始解间的局部和全局误差值进行定量计算,自适应调整策略利用该估测值调整网格剖分结构的细密度,从而提高网格单元对图案的表征效果.对人工合成图像和提花织物图像的分割实验结果表明了算法的有效性.The problem of jacquard image segmentation was discussed by approaching phase-field paradigm from a numerical approximation perspective. A modified phase-field model based on the classical phasetransition theory in the field of material science was introduced to separate and extract geometrical pattern features in jacquard images. An algorithm for numerical solving of the model was developed based on finite element approaching＇and adaptive remeshing technology. First, the above model was discretized using Pl- Lagrangian finite elements in space and semi-implicit Euler-scheme in time. Then, a mesh refinement strategy was applied to generate a mesh structure that well represented a jacquard image by adapting to its pattern content. A posteriori error estimator was presented to quantitatively estimate the local and global error between the approximation and the given image data. The adaptive remeshing scheme was driven by the posteriori error estimator to adaptively change the mesh density to improve the quality of the resulting elements in the mesh. The analysis and experimental results on synthetic and jacquard images demonstrated the effectiveness of the algorithm.

关 键 词：相位场 图像分割 提花图像 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]