Abstract: In this paper a new segmentation algorithm for textile color images is presented. First color quantization is performed on the image and a label matrix with the same size of the orignal image is generated. Second a crude segmentation is performed acording to the matrix and generates seed regions,then the pixels left through the crude segmentation are directly assigned to the seed regions. At last the initial regions are merged based on the color information and the edges of regions are smoothed.

Key words:  texture segmentation, textile image, regions merging

摘要： 提出了一种针对织物彩色图像的分割算法。该算法在颜色聚类的基础上,把每种颜色用 不同的序号来表示,生成一个与原图同大小的序号矩阵,然后根据矩阵中序号的分布进行初始分割并 产生种子区域,再把初始分割未能确定的像素指派到种子区域中,最后根据颜色信息合并过分割的区 域,并对区域边缘进行平滑,得到最终的分割结果。实验表明,该算法能较有效地实现各类彩色织物 图像的分割。

关键词: 纹理分割, 织物图像, 区域合并