基于MRA信息补偿的图像识别算法  

Image Recognition Algorithm Based on MRA Information Compensation

作　　者：王新亮[1] 黄仁[1] 

机构地区：[1]重庆大学计算机学院,重庆400044

出　　处：《计算机应用》2003年第7期89-91,共3页journal of Computer Applications

摘　　要：目前的图像识别方法对各类提取等量的特征信息,不可避免的存在过识别与不可分类问题。利用小波分解的信号自然降维特性,及多分辨率分析(MRA)空间的正交互补性,本文提出了一种通过构造信息补偿树(ICT)进行识别的新方法,实现了对各类自适应得提取不同量的特征信息,有效解决了上述问题。针对3类脸部图像进行实验,结果表明该算法是有效的且复杂性低。The present image recognition methods extract equal quantity feature information from every class,so excessive recognizing and unclassified are inevitable problems. Considering the character that wavelets decomposition naturally lows dimension and the character that multiresolution analysis spaces are orthogonal complement,this paper has proposed a new recognition algorithm by constructing information compensation tree,which can adaptively extract unequal quantity feature information from each class and has resolved the problems on above. The experiment about 3class face image is completed,and the result and analysis indicate that this algorithm is effective and has a good classification.

关 键 词：过识别 不可分类 多分辨率分析 信息补偿树 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]