计算机工程与应用 ›› 2019, Vol. 55 ›› Issue (6): 160-167.DOI: 10.3778/j.issn.1002-8331.1810-0058

• 图形图像处理 • 上一篇    下一篇

结合局部熵能量泛函与非凸正则项的图像分割

韩  明1,2,王敬涛1,孟军英1,刘教民2   

  1. 1.石家庄学院 计算机科学与工程学院,石家庄 050035
    2.燕山大学 信息科学与工程学院 河北省计算机虚拟技术与系统集成重点实验室,河北 秦皇岛 066004
  • 出版日期:2019-03-15 发布日期:2019-03-14

Energy Functional of Local Entropy Combined with Non-Convex Regularization for Image Segmentation

HAN Ming1,2, WANG Jingtao1, MENG Junying1, LIU Jiaomin2   

  1. 1.School of Computer Science and Engineering, Shijiazhuang University, Shijiazhuang 050035, China
    2.The Key Laboratory for Computer Virtual Technology and System Integration of Hebei Province, School of Information Science and Engineering, Yanshan  University, Qinhuangdao, Hebei 066004, China
  • Online:2019-03-15 Published:2019-03-14

摘要: 为了克服灰度不均匀对图像分割的影响,结合CV模型的全局能量项和LBF模型的局部能量项,引入图像局部熵信息和非凸正则项,构造新的能量泛函,提出了结合局部熵的局部能量泛函与非凸正则项的图像分割算法。该算法首先采用CV模型中的全局能量泛函得到图像的大致演化轮廓;通过构建具有局部熵信息的局部能量泛函,实现对图像的精确分割。然后,利用非凸正则项作为图像演化过程中零水平集逼近目标的又一驱动力驱动曲线演化和边缘保护。该算法利用变分水平集方法将这一新构建的能量泛函进行最小化,通过迭代更新水平集函数,完成曲线演化。最后,对比实验表明,所提出的算法可以高效、准确地分割灰度不均匀图像。

关键词: 图像分割, 局部熵, 能量泛函, 灰度不均匀, 非凸正则项

Abstract: To overcome the influence of intensity inhomogeneity on image segmentation, an image segmentation algorithm is proposed by energy functional of local entropy in combination with non-convex regularization, which combines global energy terms of CV model and local energy terms of LBF model, and local entropy information and nonconvex regularization are introduced to construct new energy functional. Firstly, global energy functional is employed to obtain coarse segmentation in CV model, which builds local energy functional based on local entropy information to obtain accurately segmentation. Secondly, non-convex regularization serves as another driving force in image evolution to drive curve evolution and preserve edge. This algorithm minimizes this newly constructed energy functional by using level set method and updates the level set function through iteration to complete the curve evolution. Finally, contrast?experimental?results?show that the proposed method?can efficiently and accurately segment intensity inhomogeneity image.

Key words: image segmentation, local entropy, energy functional, intensity inhomogeneity, non-convex regularization