基于边缘检测和区域生长的自然彩色图像分割

Natural color image segmentation based on edge detection and region growing

提出一种自然彩色图像分割方法,即利用相位一致性模型检测图像边缘,获得图像主要几何结构;并将代表区域分布的边界线作为潜在的区域模型,自动获得种子点进行区域生长实现最后分割.实验结果表明,该方法有效地克服了单一使用边缘检测和区域生长进行图像分割的缺陷,所得分割结果与人类视觉系统判断基本一致.

A new method for natural color image segmentation is proposed. The phase congruency is applied to detect edge and obtain the major geometric structures in an image. The boundary lines representing potential region distribution are taken as region model, and seeds are got automatically by boundary information to realize region segmentation by region growing. Experimental results show that this combined method can effectively avoid segmentation defection when edge detection or region growing technique is adopted separately and the segmentation results are consistent with those executed by people.