阈值分割与颜色聚类相结合的色度学图象分割方法

The Approach of Color Image Segmentation Combining Thresholding and Clustering

根据真彩色图象象素在 3D彩色空间呈现聚类分布的特点 ,可利用基于直方图的多维空间滤波技术得到目标图象 ,但会产生丢失目标信息的过度分割现象。采用颜色聚类方法改善此现象 ,在保证算法速度的基础上 ,大大提高了真彩色图象的分割效果 ,并对不同彩色模型和聚类中心有较好的鲁棒性 ,在医学染色图象处理中得到很好应用。

The pixels of a color image usually cluster in 3-D color space. According to this reason, we can use the scale space filter(SSF) method that based on thresholding histogram to get our goals. But oversegmentation may result from that. In this paper, we use the approach of color clustering to improve the result of thresholding. The proposed algorithm is very robust to different color models and centers of clustering, and its processing speed is very fast. It has been applied to divide the microscope medical image, the effect is good.