计算机工程与应用 ›› 2007, Vol. 43 ›› Issue (10): 67-69.

• 学术探讨 • 上一篇    下一篇

基于形状不变矩的t-混合模型的图像分类

张春燕 翟素兰 罗斌   

  1. 安徽大学 数学学院 清华大学软件学院
  • 收稿日期:2006-04-28 修回日期:1900-01-01 出版日期:2007-04-01 发布日期:2007-04-01
  • 通讯作者: 张春燕

Shape Classification Based on Posteriori Probability of T-mixture Model

ChunYan Zhang   

  • Received:2006-04-28 Revised:1900-01-01 Online:2007-04-01 Published:2007-04-01
  • Contact: ChunYan Zhang

摘要: 提出一种基于t-混合模型计算后验概率的统计形状分类的方法:采用Hu 形状不变矩作为图像特征,用t-混合模型的EM迭代算法,估计出模型中未知参数,计算出形状类别的后验概率。试验结果表明,该方法对每类是小样本的图像序列具有非常好的分类性质。

Abstract: This paper proposes a statistical shape classification based on the posteriori probability calculated by t-mixture model: shape feature extraction based on Hu invariant moments. We resort to EM algorithm both for the estimation of mixture components and for calculating the posteriori probability. The experimental results illuminate that approach to be suitable for small sample classification.