多阈值图像分割的遗传算法方法  

MULTI-THRESHOLDED IMAGE SEGMENTATION WITH GENETIC ALGORITHMS

作　　者：刘健庄[1] 谢维信[1] 高新波[1] 

机构地区：[1]西安电子科技大学202教研室,710071

出　　处：《模式识别与人工智能》1995年第A01期126-132,共7页Pattern Recognition and Artificial Intelligence

基　　金：国家自然科学基金

摘　　要：阈值化是图像分割中广泛使用的一种有效的工具,这些方法不难从单阈值推广到多阈值的图像分割中.但是传统的阈值化算法在求取多个阈值时需要大量的计算时间,因此影响了它们的实际使用.本文以常用的OTSU法为例提出了一种利用遗传算法来对最优阈值进行搜索的快速算法.实验结果表明,对于二阈值和三阈值的图像分割,本文方法的运算时间分别仅为传统的OTSU方法的1/4和1/100.Thresholding is widely used in image segmentation. Most of the thresholding techniques may be extended from one-thresholded segmentation to multi-thresholded segmentation. However, because the conventional algorithms require a good deal of CPU time to search optimal thresholds, their applications are restricted greatly. This paper proposes a multi-thresholded OTSU method for the search of optimal thresholds with a genetic algorithm. Our experimental results show that the proposed method requires about, 1/4 and 1/100 computational time of the conventional OTSU method in the cases of searching two thresholds and three thresholds, respectively.

关 键 词：阈值化 图像分割 遗传算法 图像识别 

分 类 号：TP391.42[自动化与计算机技术—计算机应用技术]