计算机应用 ›› 2012, Vol. 32 ›› Issue (10): 2827-2831.DOI: 10.3724/SP.J.1087.2012.02827

• 图形图像处理 • 上一篇    下一篇

小空间占用的快速彩色图像特征抽取方法

罗婵娟1,2,3,朱嘉钢1,2,3,陆晓3   

  1. 1. 江南大学 物联网工程学院,江苏 无锡 214122
    2. 江南大学 物联网应用技术教育部工程研究中心,江苏 无锡 214122
    3. 江苏晓山信息产业股份有限公司,江苏 无锡 214122
  • 收稿日期:2012-04-17 修回日期:2012-05-28 发布日期:2012-10-23 出版日期:2012-10-01
  • 通讯作者: 罗婵娟
  • 作者简介:罗婵娟(1985-),女,陕西宝鸡人,硕士研究生,主要研究方向:模式识别与人工智能、软件工程;朱嘉钢(1957-),男,上海人,副教授,博士,主要研究方向:人工智能与模式识别、软件工程;陆晓(1961-),男,江苏无锡人,工程师,主要研究方向:物联网工程。
  • 基金资助:
    国家自然科学基金资助项目;江苏省自然科学基金项目

Color image feature extracting method with small space occupying and fast speed

LUO Chan-juan1,2,3,ZHU Jia-gang1,2,3,LU Xiao2   

  1. 1. Engineering Research Center of Internet of Things Applied Technology,Ministry of Education,Jiangnan University,Wuxi Jiangsu 214122,China
    2. Jiangsu Xiaoshan Information Industry Limited, Wuxi Jiangsu 214122,China
    3. School of IOT, Jiangnan University, Wuxi Jiangsu 214122,China
  • Received:2012-04-17 Revised:2012-05-28 Online:2012-10-23 Published:2012-10-01
  • Contact: LUO Chan-juan

摘要: 为了有效降低已有彩色图像特征抽取算法的空间占用,使得这类算法可以适用于仅有有限计算能力和计算空间的计算环境,提出了一种小空间占用的快速彩色图像特征抽取方法。此方法首先用无迭代双边二维主成分分析方法NIB2DPCA对彩色图像的R、G、B三个通道分别做特征抽取;然后把抽取到的三个特征矩阵重构为一个二维矩阵;接着用NIB2DPCA对此二维矩阵抽取特征得到最终的分类特征矩阵。最后用最近邻分类器验证提出方法的有效性。在CVL和FEI人脸库上的大量实验表明,提出的方法采用两次特征抽取方法对彩色图像的信息进行了有效的压缩从而使计算过程中占用的内存空间减小了两个数量级以上,由此导致了计算时间的缩短,计算速度的提高;而且识别率还有所提高。

关键词: 彩色图像识别, 特征抽取, 无迭代双边二维主成分分析, 二维主成分分析, 主成分分析

Abstract: A new method for color image feature extraction was proposed in order to effectively reduce the space occupying of the existing similar algorithms, so that such type of algorithms can be used in the computing environment with only limited computing ability and space. First of all, Non-Iteration Bilateral projection based Two Dimensional Principal Component Analysis (NIB2DPCA) was employed to extract feature information from three channels of a given color image respectively. Then the three pre-extraction feature matrices of the color image was reconstructed to form a two dimensional matrix. After that, NIB2DPCA was again employed to extract features of the matrix to obtain the final features. Finally,the nearest neighbor classification was employed to verify the performance of the method. A large number of experimental results on CVL and FEI face databases show that the color image data is efficiently compressed so the memory space occupying is reduced by more than two orders of magnitude as the result of the twice feature extraction. The calculation time is largely reduced and the calculation speed is largely improved due to the reduced memory space occupying, while the recognition rate is still significantly increased.

Key words: color image recognition, feature extraction, Non-Iteration Bilateral Projection Based 2DPCA (NIB2DPCA), Two Dimensional Principal Component Analysis (2DPCA), Principal Component Analysis (PCA)

中图分类号: