NSCT和CS-LBP的低分辨率人脸识别  

Low-resolution face recognition based on NSCT and CS-LBP

作　　者：孙万春 张建勋[1] 郑集元 陈虹伶 朱佳宝 SUN Wan-chun;ZHANG Jian-xun;ZHENG Ji-yuan;CHEN Hong-ling;ZHU Jia-bao(College of Computer Science and Engineering, Chongqing University of Technology, Chongqing 400054, China)

机构地区：[1]重庆理工大学计算机科学与工程学院,重庆400054

出　　处：《计算机工程与设计》2018年第12期3823-3828,共6页Computer Engineering and Design

基　　金：四川省高校重点实验室开放基金项目(2015WZJ02)

摘　　要：针对目前监控视频采集的图像人脸信息分辨率较低问题,提出一种基于非下采样Contourlet变换(NSCT)和分块中心对称局部二进制模式(CS-LBP)的低分辨率人脸识别方法。利用NSCT在方向和尺度上变换去分解过滤分辨率不高的图像;借助CS-LBP在光照变换时受影响较小的优点,分块提取经过分解得到的高低频图像信息,将得到的特征直方图级联起来,获取较为完整的能表达人脸的向量直方图。针对最后图像维度高,使用PCA主成分分析法降维,提高实时性能。实验结果表明,该方法在低分辨率人脸识别环境下能够较好提取人脸特征,获取更高的识别率。Aiming at the low resolution of the current surveillance video,leading to low face recognition rate,a low resolution face recognition method based on nonsubsampled Contourlet transform(NSCT)and center-symmetric local binary pattern(CS-LBP)was proposed.NSCT was used to transform and decompose the image with low resolution in the direction and scale with the advantage of CS-LBP being less affected during illumination conversion.The high and low frequency image information obtained by decomposition was extracted by block and feature histograms were cascaded to obtain a more complete vector histogram that expressed faces.For the final image dimension,PCA principal component analysis was used to reduce dimensionality and improve real-time performance.Experimental results show that the proposed method can extract facial features better and achieve higher recognition rate in low-resolution face recognition environment.

关 键 词：低分辨率 人脸识别 非下采样CONTOURLET变换 特征提取 中心对称局部二进制模式 主成分分析 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]