计算机应用 ›› 2013, Vol. 33 ›› Issue (03): 670-673.DOI: 10.3724/SP.J.1087.2013.00670

• 多媒体处理技术 • 上一篇    下一篇

融合视觉模型和最大类间方差的阈值分割算法

邹小林1,冯国灿2*   

  1. 1.肇庆学院 数学与信息科学学院,广东 肇庆 526061;
    2.中山大学 数学与计算科学学院, 广州 510275
  • 收稿日期:2012-09-28 修回日期:2012-11-01 出版日期:2013-03-01 发布日期:2013-03-01
  • 通讯作者: 邹小林
  • 作者简介:邹小林(1975-),男,湖南衡阳人,讲师,博士,主要研究方向:图像处理、模式识别; 冯国灿(1962-),男,湖北孝感人,教授,博士,主要研究方向:计算机视觉、模式识别、图像处理。
  • 基金资助:

    国家自然科学基金资助项目(60975083, 61272338)。

Image thresholding segmentation based on human vision model and maximum between-cluster variance

ZOU Xiaolin1, FENG Guocan2*   

  1. 1.School of Mathematics and Information Sciences, Zhaoqing University, Zhaoqing Guangdong 526061, China;
    2. School of Mathematics and Computational Science, Sun Yat-sen University, Guangzhou Guangdong 510275, China
  • Received:2012-09-28 Revised:2012-11-01 Online:2013-03-01 Published:2013-03-01
  • Contact: ZOU Xiao-lin

摘要: 针对传统二维直方图的区域划分方法存在把图像的部分目标点和背景点错误划分为边缘点或噪声点,而把部分边缘点和噪声点划分为目标点和背景点的缺点,以及传统二维最大类间方差阈值分割算法的时间复杂度较高的缺点,提出了采用视觉模型构造二维直方图,并提出了该二维直方图的区域划分方法,同时还把提出的二维直方图应用到最大类间方差阈值分割算法中。根据分割时间、分类误差、均匀性等定量评价标准,做了一系列实验,与几种典型的二维阈值分割算法相比,提出的阈值分割算法在降低计算复杂度的同时还具有很好的分割性能。

关键词: 人类视觉模型, 图像分割, 阈值选取, 最大类间方差法

Abstract: Because the area division of the traditional two-dimensional histogram has such shortcomings as part of the target points and background points are divided into edge points or noise points, while part of the edge points and noise points are divided into the target point and background points and the traditional two-dimensional image thresholding segmentation algorithms' time complexity are high, a new two-dimensional histogram was proposed by using human vision model, and a new region division method was proposed to the proposed histogram, and at the same time, the proposed histogram was applied to image thresholding segmentation based on maximum between-cluster variance. According to segmentation time, misclassification error and uniformity, a series of experiments were carried out to show the proposed algorithm reduces the time complexity and has good segmentation performance compared with several typical two-dimensional threshold segmentation algorithms.

Key words: human visual model, image segmentation, threshold selection, maximum between-cluster variance

中图分类号: