电子学报 ›› 2015, Vol. 43 ›› Issue (3): 523-528.DOI: 10.3969/j.issn.0372-2112.2015.03.017

• 学术论文 • 上一篇    下一篇

基于Gabor特征和字典学习的高斯混合稀疏表示图像识别

詹曙1, 王俊1, 杨福猛2, 方琪1   

  1. 1. 合肥工业大学计算机与信息学院, 安徽合肥 230009;
    2. 三江学院电子信息工程学院, 江苏南京 210012
  • 收稿日期:2013-12-03 修回日期:2014-08-29 出版日期:2015-03-25
    • 作者简介:
    • 詹 曙 男,1968年生于安徽合肥.副教授、硕士生导师,合肥工业大学计算机与信息学院.研究方向为三维人脸识别和医学图像处理. E-mail:shu_zhan@hfut.edu.cn 王 俊 男,1989年生于安徽合肥.硕士研究生,合肥工业大学计算机与信息学院.研究方向为数字图像处理和模式识别. E-mail:wangjunhfut@163.com;杨福猛 男,1968年生于安徽和县,副教授,三江学院电子信息工程学院.研究方向:电子与通讯,图像及信号处理等.
    • 基金资助:
    • 国家自然科学基金 (No.61371156); 安徽省科技攻关计划 (No.1401B042019)

Gaussian Mixture Sparse Representation for Image Recognition Based on Gabor Features and Dictionary Learning

ZHAN Shu1, WANG Jun1, YANG Fu-meng2, FANG Qi1   

  1. 1. School of Computer & Information, Hefei University of Technology, Hefei, Anhui 230009, China;
    2. School of Electronic Information Engineering, Sanjiang University, Nanjing, Jiangsu 210012, China
  • Received:2013-12-03 Revised:2014-08-29 Online:2015-03-25 Published:2015-03-25

摘要:

为了克服图像识别中光照,姿态等变化带来的识别困难,同时提高稀疏表示图像识别的鲁棒性,本文提出了一种基于Gabor特征和字典学习的高斯混合稀疏表示图像识别算法.高斯混合稀疏表示是基于最大似然估计准则,将稀疏保真度表示为余项的最大似然函数,最终识别问题转化为求解加权范数的优化逼近问题.本文算法首先提取图像的Gabor特征;然后对Gabor特征集进行字典学习,由于在学习过程中引入了Fisher准则作为约束,学习得到具有类别标签的新字典;最后使用高斯混合稀疏表示识别方法进行分类识别.在3个公开数据库(人脸数据库AR库和FERET库以及USPS手写数字库)上的实验结果验证了该算法的有效性和鲁棒性.

关键词: Gabor特征, 稀疏表示, fisher字典学习, 最大似然估计

Abstract:

To overcome the problems of the illumination and pose variations in image recognition, the algorithm of Gaussian mixture sparse representation for image recognition based on dictionary learning and Gabor features is proposed.Based on the maximum likelihood estimation principle, a mixture Gaussian sparse coding model is proposed to express the discriminating items to the maximum likelihood function of residuals, so the problem of identification is converted to the optimal weighted norm approximation problem.This approach extracts the Gabor features of the images by the Gabor filter, and then uses the Gabor features to learn a new dictionary.As the Fisher criterion is added in the learning process as a constraint, a new dictionary with category labels can be obtained.Finally, the method of Gaussian mixture sparse representation is used for classification and identification.The experimental results in three public databases demonstrate that the algorithm proposed is effective and robust.

Key words: Gabor features, sparse representation, fisher dictionary learning, maximum likelihood estimation

中图分类号: