Abstract: This paper puts forward a new segmentation method for color image.Image’s color feature is extracted from HSI model at first，the second，image’s texton feature is extracted from color image’s intensity，this paper achieves this goal by the Gabor transform irrelevant to orientation parameters，after these two steps，this paper gets a 3 dimensions color-texture feature vector，at last.It is used Kernel Fuzzy Cluster Method（KFCM） as segment algorithm.Compared with other method，the strategy has better performance，experimental results indicates that the method has more accurate than that only use color or texture.

Key words: color image, segmentation, color-texture, Kernel Fuzzy Cluster Method（KFCM）

摘要： 提出了一种利用色彩-纹理综合特征进行彩色图像分割的新方法。首先利用HSI色彩模型提取图像的色彩信息，其次，采用与方向无关的Gabor变换对彩色图像的强度信息进行处理，提取图像的纹理基元。根据上述过程所获得的色彩-纹理特征，采用核模糊聚类方法实现彩色图像的自动分割。实验结果表明，所提出的方法综合利用了图像的色彩、纹理信息，使彩色图像的分割结果更准确。

关键词: 彩色图像, 分割, 色彩纹理, 核模糊聚类