基于DWT和nova-form聚类法的医学图像颜色特征提取

Extracting Color Characters Based on DWT and an New Cluster Arithmetic-Nova form

提出一种基于小波变换的医学图像颜色特征提取新方案。根据小波变换后各子带所含的图像信息量调整主色提取种数。对低频子带,首次提出了一种适用于主色提取的无监督聚类分析方法——nova-form;而对高频子带则采用k-medoids算法。实验证明,nova-form聚类算法较经典算法有更好的聚类效果,新方案在检索速度和检索精度上能较好地兼顾。

A new technology in extracting color characters of images based on wavelet-transform is advanced. As to low frequency, for the first time, a non-supervise clustering method is put forward, which is the same with extracting major colors; and as to high frequency, k-medoids is used. The nova-form, the new arithmetic, has the better clustering effect comparing to other classic arithmetic as proved. As well, the advanced way of pickup color characters can give preferably attentions to both precision and efficiency of searches.