作者投稿和查稿 主编审稿 专家审稿 编委审稿 远程编辑

计算机工程 ›› 2012, Vol. 38 ›› Issue (04): 179-181. doi: 10.3969/j.issn.1000-3428.2012.04.058

• 人工智能及识别技术 • 上一篇    下一篇

改进的自商图算法

胡 华   

  1. (枣庄学院信息科学与工程学院,山东 枣庄 277100)
  • 收稿日期:2011-08-30 出版日期:2012-02-20 发布日期:2012-02-20
  • 作者简介:胡 华(1977-),女,讲师、硕士,主研方向:图像处理

Improved Self Quotient Image Algorithm

HU Hua   

  1. (College of Information Science and Engineering, Zaozhuang University, Zaozhuang 277100, China)
  • Received:2011-08-30 Online:2012-02-20 Published:2012-02-20

摘要: 针对人脸识别中的光照变化问题,提出一种改进的自商图算法。对光照图像进行伽玛变换,使用非下采样轮廓波变换对图像进行多尺度多方向分析,对各方向子带进行Wiener滤波,利用自商图模型提取人脸图像的光照不变特性。Yale B与CMU PIE人脸库上的实验结果表明,与传统算法相比,该算法的平均识别率更高。

关键词: 人脸识别, 自商图像, 非下采样Contourlet变换, Wiener滤波, 伽玛变换

Abstract: In order to eliminate the effect of illumination on face recognition, an improved Self Quotient Image(SQI) algorithm is proposed. The algorithm performs nonlinear transform Gamma correction on image under various lighting conditions. Nonsubsampled Contourlet Transform (NSCT) is used for analysis with multiscale and multidirection, after that Wiener filter is applied to high frequency directional subbands for illumination invariant extraction. Experimental results on Yale B and CMU PIE databases show that the algorithm can effectively eliminate the effect of illumination on face recognition.

Key words: face recognition, Self Quotient Image(SQI), Nonsubsampled Contourlet Transform(NSCT), Wiener filtering, Gamma transform

中图分类号: