基于遗传算法的彩色图像多阈值分割方法研究  

Study on multi-threshold segmentation for color image based on genetic algorithm

作　　者：张国权[1,2] 

机构地区：[1]兰州理工大学电信学院,甘肃兰州730050 [2]兰州教育学院信息科学系,甘肃兰州730020

出　　处：《电子设计工程》2011年第9期43-45,共3页Electronic Design Engineering

摘　　要：针对彩色图像多阈值分割中阈值个数自动确定困难和优化时间长的问题,首先提出一种新的HSV空间中彩色图像投影预处理方法,然后计算待分割图像的颜色粗糙度,并根据颜色粗糙度确定图像分割的阈值数量,为提高分割效率,利用遗传算法搜索最优分割阈值组合,为提高分割精度,在遗传算法适应度函数设计时既考虑了类内的颜色离散度,又考虑了像素的空间关系。实验结果表明该方法具有较为理想的分割效果,对光线变化具有较好的鲁棒性。Aiming at the difficulty of threshold number calculation and optimization, a method for color image pre-processing in HSV space is proposed, and image＇s rough degree is calculated through color distance. The number of threshold for image segmentation is calculated according to roughness ; then the genetic algorithm is utilized to find the most reasonable threshold combination ,when the fitness function is designed , both information on color and space are all be considered. Experiments prove that this method has excellent effect on color image segmentation and has robustness against illumination variation.

关 键 词：HSV颜色空间 粗糙度 遗传算法 多阈值 图像分割 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]