您的位置:首页  > 论文页面

人脸识别的相对梯度方向边缘幅值模式

发表时间:2014-02-28  浏览量:1760  下载量:677
全部作者: 杨利平,叶洪伟
作者单位: 重庆大学光电工程学院
摘 要: 针对方向边缘幅值模式(pattern of oriented edge magnitude,POEM)在剧烈光照变化情况下无法获得足够特征描述信息的问题,在分析相对梯度幅值图像特点的基础上,结合子空间流形学习,提出相对梯度方向边缘幅值模式(relative gradient patterns of oriented edge magnitudes,RGPOEM)特征描述方法。该方法根据图像的梯度方向对相对梯度幅值图像进行分解、滤波、局部二值模式(local binary patterns,LBP)编码和保局投影(locality preserving projections,LPP)降维,形成低维的RGPOEM特征。在Yale和PIE子集上的人脸识别实验证实,RGPOEM继承了POEM特征描述的优点,同时对光照变化具有更好的健壮性。
关 键 词: 模式识别;人脸识别;特征描述;相对梯度;保局投影
Title: Face recognition using relative gradient patterns of oriented edge magnitudes
Author: YANG Liping, YE Hongwei
Organization: College of Optoelectronic Engineering, Chongqing University
Abstract: When illumination condition changes drastically, it is difficult for pattern of oriented edge magnitude (POEM) method to acquire enough feature description information. To alleviate the issue, a relative gradient pattern of oriented edge magnitudes (RGPOEM) method is proposed in this paper. The method decomposes the relative gradient magnitude image into several sub images according to the orientations of gradient. Each of these sub images is then filtered and encoded using local binary patterns. Finally, all the encoded local binary patterns (LBP) histogram features are connected by lexicographic ordering and reduced to a low-dimensional subspace using locality preserving projections (LPP) to form a low-dimensional RGPOEM feature. Experimental results on Yale and PIE subsets indicate that RGPOEM, which inherits the superior characteristics of POEM, exhibits much robustness to illumination variance.
Key words: pattern recognition; face recognition; feature description; relative gradient; locality preserving projections
发表期数: 2014年2月第4期
引用格式: 杨利平,叶洪伟. 人脸识别的相对梯度方向边缘幅值模式[J]. 中国科技论文在线精品论文,2014,7(4):300-305.
 
0 评论数 0
暂无评论
友情链接