基于改进K-均值聚类的纸币冠字号图像分割算法

Image segmentation of banknotes numbers based on improved K-means clustering algorithm

针对传统彩色图像分割方法的局限性,提出了基于HSI色彩空间和改进K-均值聚类的图像分割方法,通过将彩色图像分解成三个相互独立的H、S、I分量,利用各个分量特点及其直方图确定聚类类别和初始聚类中心,在高饱和度区和低饱和度区分别聚类,并将聚类结果合并取交集,从而分割出目标区域。将该方法用于纸币冠字号码图像分割,经仿真验证,结果不受噪声和局部边缘变化的影响,分割效果得到明显提升,为后续冠字号准确识别提供了良好的基础。

To overcome the limitations of traditional image segmentation method, a new method based on HSI colour space and improved K-means clustering algorithm is proposed. In HSI space, the colour image is de- composed into three independent components. By using the features and histograms of these components, the clustering centres and categories are obtained. Then the clustering is implemented in high saturation area and low saturation area respectively, after the results are combined and mixed, the object area will be extracted. As an exam, the provided method is used to extract the banknotes numbers, the results show that the method is effective and not influenced by noise and edge change, it provides a good foundation for the accurate identi- fication of banknotes numbers.