摘要
为了有效地对受噪声影响的图象进行分析 ,提出了两种基于二维遗传算法的图象动态分割算法 .在这些算法中 :1 )分别采用了以阈值曲面和模糊隶属度曲面为染色体的二维染色体编码方式 ;2 )采用了全局阈值化算法和模糊集合理论初始化种群 ;3)采用 Hopfield网络的能量函数形式 ,结合 FCM算法和现有阈值化算法中的一般性分割准则构造适应度函数 .利用实际图象将所提出的算法与一些典型算法进行了分割比较实验 ,结果表明所提算法有较好的抗噪效果 .
To effectively accomplish analysis tasks for noisy images, two new algorithms for dynamic image segmentation based on 2 D genetic algorithms are proposed. In these new algorithms:(a)coding with 2 D chromosome by taking respectively the surface of thresholds and surface of fuzzy membership is adopted; (b)initialization of population with global thresholding methods and fuzzy set theory is experienced; (c)construction of fitness function by using energy function of Hopfield neural network as well as by integrating objective function of FCM algorithm, thresholding algorithms and general criteria of image segmentation is employed. A number of experiments using real images are performed, in which the proposed algorithms are compared with some typical segmentation algorithms. Results show that the new algorithms have superior performance for noisy images.
出处
《自动化学报》
EI
CSCD
北大核心
2000年第5期685-689,共5页
Acta Automatica Sinica
基金
国家自然科学基金资助项目 !( N o.6 96 72 0 2 9)
关键词
二维遗传算法
图象动态分割
图象分析
计算机
2-D genetic algorithms, fitness function, dynamic image segmentation.
