• 中国计算机学会会刊
  • 中国科技核心期刊
  • 中文核心期刊

计算机工程与科学

• 论文 • 上一篇    下一篇

抑制纹理信息的偏置场变分图像分割模型

李虎,汪西莉   

  1. (陕西师范大学计算机科学学院,陕西 西安 710119)
  • 收稿日期:2015-11-20 修回日期:2016-01-11 出版日期:2017-02-25 发布日期:2017-02-25
  • 基金资助:

    国家自然科学基金(41171338,41471280,61401265);陕西省自然科学基础青年项目(2014JQ8312 )

A bias field variational image segmentation
 model restraining texture information

LI Hu,WANG Xi-li   

  1. (School of Computer Science,Shaanxi Normal University,Xi’an 710119,China)
  • Received:2015-11-20 Revised:2016-01-11 Online:2017-02-25 Published:2017-02-25

摘要:

偏置场变分水平集图像分割模型利用原始图像的局部灰度信息,可以对灰度不均匀图像进行有效的分割,但当灰度图像中存在纹理时,分割效果往往很差。针对这一问题,提出抑制纹理信息的偏置场变分水平集图像分割模型。利用一种基于纹理几何结构的纹理描述符描述图像中不同的纹理区域,使得不同纹理区域对比更加明显,相同纹理区域更加平滑,通过抑制纹理信息使后续的图像分割在纹理部分的错分大大减少。实验结果表明,相比偏置场变分模型,所提模型对自然及人工合成纹理图像均获得更好的分割结果。
 

关键词: 图像分割, 偏置场, 变分模型, 纹理描述符, 纹理图像

Abstract:

The variational level set image segmentation based on the bias field can segment intensity inhomogeneity images using the local image information. However, the model cannot do it well when there are textures in the image. To solve the above problem, we propose a bias field variational level set image segmentation model to suppress texture information. The texture can be restrained by the intrinsic texture descriptor based on the texture geometric structure and it can enhance the contrast among different texture regions and smooth the image in the same texture region. It can reduce the mistakes of texture region segmentation by restraining the texture. Experimental results indicate that the proposed model can better segment natural and synthetic texture images in comparison with the bias field variational model.
 

Key words: image segmentation, bias field, variational model, texture descriptor, texture image