摘要
针对传统医学图像分割算法时间复杂度高、分割精度低等问题,提出一种基于一维Otsu的自动多阈值分割算法.考虑到医学图像信息的复杂性,引入基于梯度、灰度、距离的综合信息直方图替代传统的灰度直方图,并分别赋予这3个信息相应的权值.采用kd-树作为框架快速自动确定阈值个数,进而实现Otsu对医学图像的自动多阈值分割.与最大熵、基于粒子群优化的Otsu算法等进行对比实验的结果表明,该算法的分割性能优于其他算法.
Aiming at the problems of high time complexity and low accuracy of traditional medical image segmentation algorithm,we proposed a segmentation algorithm based on one-dimensional Otsu automatic multiple threshold. Taking into account the complexity of medical image information, we introduced a comprehensive information histogram based on gradient,gray scale and distance to replace the traditional gray histogram,and gave the corresponding weights of the 3 information respectively. Kd-tree was used as a framework to quickly determine the number of threshold automatically,and then to realize the automatic threshold segmentation of medical images by Otsu. Compared with the maximum entropy and Otsu based on particle swarm optimization algorithm,experimental results show that the segmentation performance of the proposed algorithm is superior to other algorithms.
出处
《吉林大学学报(理学版)》
CAS
CSCD
北大核心
2016年第2期344-348,共5页
Journal of Jilin University:Science Edition
基金
国家自然科学基金青年基金(批准号:61305046)
吉林省自然科学基金(批准号:20140101193JC)
吉林省青年科学基金(批准号:20130522117JH)