双曲线二维Otsu阈值分割算法  

Hyperbolic Two Dimensional Otsu Threshold Segmentation Algorithm

作　　者：赵恒 安维胜[1] 杨陶[1] ZHAO Heng;AN Weisheng;YANG Tao(College of Mechanical Engineering,Southwest Jiao Tong University,Chengdu 610031)

机构地区：[1]西南交通大学机械工程学院

出　　处：《计算机与数字工程》2019年第8期2033-2038,共6页Computer & Digital Engineering

摘　　要：由于二维直方图阈值分割算法存在的区域错分导致分割质量和抗噪性能的不足,提出一种新的自适应区域划分阈值图像分割算法。采用两条曲线划分二维直方图,获得符合图像信息分布的区域划分结果。该算法用边缘检测算子获得图像边缘信息,再用三次样条曲线拟合法拟合边缘得到两条曲线,将所有像素划分为噪声和目标背景两类,对噪声灰度值进行替换;用改进的二维Otsu算法对噪声处理后的图像进行阈值分割。与改进的传统二维直方图分割算法和改进的二维直方图斜分算法的实验结果对比表明,论文算法运行效率高,且具有更好的分割质量,更强的抗噪能力和自适应能力。Since a poor segmentation quality,anti-noise capacity of(2D)Otsu algorithm with region misclassified,a novel Otsu adaptive image segmentation algorithm based an improved region division is proposed.The region is divided by two curves,which is consistent with the image information distribution.The proposed algorithm applies contour detector to obtain edges and use three spline fitting method to fit edges.Then,all pixels are converted into noise and target,and all gray values of noise are replaced.Finally,the improved 2D Otsu segmentation method is utilized to separate object from background after the gray values of noise were replaced.Compared with the improved traditional 2D histogram threshold segmentation algorithm and improved 2D histogram oblique segmentation algorithm,experimental results show that the novel algorithm is efficient,and has better segmentation quality,stronger anti-noise capacity and adaptive ability.

关 键 词：图像分割 边缘检测 区域划分 曲线拟合 二维直方图 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]