基于核的密度函数聚类的彩色图像分割方法

Color Image Segmentation Based on Kernelized Density Function Clustering Algorithm

提出一种基于核方法的密度函数聚类方法和小波变换的快速彩色图像分割方法。对密度函数聚类方法改进,通过引入核方法生成基于核的密度函数聚类,用于彩色图像聚类数目上限和初始聚类中心;利用小波变换的多分辨率特性,在分辨率最大的子带进行聚类数目的确定以减少计算量,然后把分割结果逐层延伸到原始尺寸图像得到最终分割结果。

The paper proposes a color image Segmentation algorithm based on Kernelized density function clustering algorithm and wavelet transform. A Kernelized density function clustering algorithm is presented by combining kernel function and density function clustering algorithm to define the optimal cluster number and initialize cluster centers; By using the multi-resolution of wavelet transform, defining the optimal cluster number is implemented on the sub-band of maximum size to reduce computing load and the coarse segmentation is used to initialize the next level until the initial image.