基于形状信息的Bayes分类方法

A Bayes Classification Algorithm Based on Shape Information

本文提出了一种新的基于形状信息的Bayes分类方法,以实现对图像中单个物体的分类。该方法首先运用图像边缘提取和配准算法,构造一个形状相似性能量泛函,并利用其计算形状信息的先验概率。然后,结合图像中物体其它特征的后验概率,通过Bayes方法进行分类。本文将该方法应用于一个病原菌图像分类的实际问题,实验结果表明,该方法是十分有效的,不仅降低了分类所需的特征维数,而且提高了分类精度,并能满足实际问题中所要求的计算速度。

In this paper,a new Bayes classification algorithm based on the shape information is proposed to classify the objects in the image.In this method,an energy functional which indicates the similarity of different shapes is introduced to calculate the prior probability of the shape information by applying the image edge extraction and image registration algorithms,and then the objects are clustered by the Bayes method with some posterior probabilities of other features.The presented algorithm has been applied to a practical pathogeny bacteria image classification problem,and the experimental results show the high effciency of our algorithm,which not only reduces the feature dimensions of samples,but also improves the classification accuracy.Moreover,it can fulfill the requirement of computing speed in the practical problem.