Computer Engineering and Applications ›› 2008, Vol. 44 ›› Issue (5): 70-72.

• 学术探讨 • Previous Articles     Next Articles

Feature fusion based on 2DPCA and its application

WEI Xu,TAO Bing-jie   

  1. EE of UESTC,Chengdu 610054,China
  • Received:2007-06-08 Revised:2007-08-21 Online:2008-02-11 Published:2008-02-11
  • Contact: WEI Xu

基于2DPCA的特征融合方法及其应用

魏 旭,陶冰洁   

  1. 电子科技大学 电子工程学院,成都 610054
  • 通讯作者: 魏 旭

Abstract: From researching on the universal principle of feature fusion of image,a new algorithm was proposed which based on the 2 dimension principal component analyses(for short 2DPCA).First,choose the multi-feature of target image which includes fractal feature,multi-direction and multi-scale gradient feature,local gray probability feature to construct a characteristic vector.Get a transforming matrix from using 2 dimension principal component analyses to the characteristic vector.And multiply the transforming matrix and characteristic vector to get a new syntheses feature.This syntheses feature is the feature from fusion.Simulation on image matching of small-weak target shows that this algorithm is more effective than conventional methods of matching based on gray and methods based on PCA,and has minor difference of time to the last method.

Key words: 2DPCA, feature fusion, small-weak target, image matching

摘要: 通过对图像特征融合的一般规律的研究,提出了一种基于二维主成分分析(简称2DPCA)的图像特征融合算法。首先选取包括分形特征、多向多尺度梯度特征、局域灰度概率特征在内的目标图像的多种特征,组成特征向量,对该向量进行二维主成分分析,得到一个变换矩阵,再利用该变换矩阵和原特征向量的乘积得到新的综合特征。该综合特征即为经过融合后得到的特征。在对弱小目标匹配跟踪的仿真结果表明,该方法效果优于常规的灰度匹配和基于PCA的特征融合方法,且耗时与后者相差不大。

关键词: 二维主成分分析, 特征融合, 弱小目标, 图像匹配