摘要
提出了1种基于自组织神经网络的彩色图像分割方法.它首先在空间-颜色联合域内利用自组织神经网络对图像进行聚类,再结合区域生长的方法和区域邻接列表,对图像中的相似颜色区域进行合并,并去除像素数量较少的小区域,从而得到最后的分割结果.在聚类后的图像上进行分割,可以较好的把1个物体对象只表示为1个或少数几个分割区域,明显减少对象物数目,较好地克服了传统区域生长方法的过分割问题.
A new method for color image segmentation based on self-organized neural network is proposed. Firstly, color cluster in spatial-color domain is made by use of the method of self-organizing neural network. Secondly, the method of region growing and region adjacency list is used to merge similar regions and eliminate the small regions. Then, the final segmentation result is abtained. By the image segmentation after color cluster, an object can be segmented into one or several regions, and the number of regions can be reduced obviously.
出处
《山东大学学报（工学版）》
CAS
2005年第6期67-70,共4页
Journal of Shandong University（Engineering Science）
基金
国家自然科学基金资助项目(编号:60104009)
关键词
图像分割
区域邻接矩阵
自组织神经网络
空间-颜色域
image segnentation
region adjacency matrix
self-organizing neural network
spatial-color domain
