基于最优家族遗传算法的改进二维熵图像分割  

Reformative 2-Demensional Entropy Method for Image Segmentation Based on Optimum Family Genetic Algorithm

作　　者：徐晓军[1] 李建华[1] 王孙安[1] 郭咏虹[1] 

机构地区：[1]西安交通大学机械工程学院,西安710049

出　　处：《西安交通大学学报》2004年第11期1182-1185,共4页Journal of Xi'an Jiaotong University

基　　金：陕西省自然科学基金资助项目 (2 0 0 1X1 7) ;陕西省机械制造装备重点实验室资助项目 (0 3IF0 6)

摘　　要：分析了最大二维熵图像分割算法的基本原理,提出了一种改进的二维熵图像分割算法.该算法同时考虑了孤立像素点的灰度信息和像素点的空间相关性,并对目标的边缘进行检测,因此保留了更多的图像边缘信息.引入搜索区域、群体规模可变的最优家族遗传算法对阈值进行搜索,不仅提高了算法的搜索速度,而且避免了早熟现象.实验结果表明,分割256×256的Lena图100次,平均时间为1 5937s,平均进化代数为2 5037,且边缘信息得到了很好保留.改进算法在分割速度和分割精度上比普通二维熵的分割算法有显著的提高,说明了算法的有效性.Based on the principle of 2-demensional entropy, an improved algorithm for image segmentation is put forward, where the spatial correlation and grey scale are considered, and the target edgy is detected to kept down more edgy information. The threshold is searched by optimum family genetic algorithm (OF-GA) which has the ability to change the search space and population size. Compared with the other similar segmentation algorithm, the newly proposed one facilitates accelerating the computing rate, improving the accuracy and preventing the prematurity. To illustrate the algorithm validity, the Lena image whose size is 256X256 pixels is segmented 100 times. The results show that the average time reaches to 1.5937 s, the average evolution number approaches to 2.5037, and the edgy informations are kept down perfectly.

关 键 词：二维熵 图像分割 遗传算法 边缘检测 灰度 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]