多分辨率分析的人脸识别

VIP免费

3.0 侯斌 2024-11-19 5 4 660.65KB 45 页 15积分

侵权投诉

摘要

随着计算机互联网技术的发展，如何快速地对大量的人脸图像进行识别，成

为近四十多年来的研究热点。因此，人脸识别技术获得了飞速的发展，并在国家

反恐、信息安全、访问控制等领域有着广泛的应用。但是研究人员发现，人脸识

别仅仅在约束条件下能够达到满意的效果，而在非约束条件下的人脸识别还远未

成熟。

在人脸识别研究中，如何快速对获得的人脸图像进行识别，并且降低外界条

件对于人脸识别的影响，是人脸识别的研究热点之一。

本文首先广泛研究了人脸识别的发展历史和研究现状，对人脸识别系统的组

成结构、公共资源以及当前的研究热点和难点进行了深入研究。最终将研究重点

集中于提高人脸识别速度方面。

为提高人脸识别的速度，论文提出了一种基于分类的图像检索算法。算法以

图像间的相关系数作为距离进行聚类，并确定每类图像的中心图像。图像检索过

程分为两步：首先寻找匹配程度最高的中心图像，然后在该中心图像所在类中寻

找最佳匹配图像。因为分类和确定类中心图像均可离线操作，所以算法有效缩短

了图像检索时间。实验仿真证明，算法有效。

论文还提出了基于多分辨率分析和分类的人脸识别改进算法。算法首先利用

小波变换对训练样本进行多尺度小波分解，以方差最大的小波系数间相关系数作

为分类距离，对样本进行分类，并确定每类图像的中心图像。人脸识别算法类似

于前面的识别算法。因为算法对原始人脸图像做了小波变换，并提取图像的一部

分小波系数，所以达到了降维目的，减少了计算量，有效提高了人脸识别的速度。

经过仿真实验证明，论文提出的算法有效。

关键词：人脸识别多分辨率分析图像检索相关系数小波图像分类

信息熵

ABSTRACT

With the development of the Internet Technology , the active research area is how

to recognize the face image quickly in recent forty years. So the technology of the face

recognition has a fast development . Although the face recognition have a broad

application in the national counter-terrorism , information security , access control and

other areas , it can not perform the recognition task successfully under practical

conditions . That is , it is satisfactory under ideal environment .

How to fast recognize the face image and reduce the influence of the practical

environment is one of the active research problems.

The paper makes a comprehensive history research summary of face recognition

firstly and researches some difficulty problems as the structure of face recognition

system and public resources. Then the paper focuses its research aspect on accelerating

face recognition speed.

In order to increase the recognition speed, an efficient algorithm which based on

the image classification is being used to the image retrieval . The cross correlation

among the images is being used as the distance to cluster , and the central image of per

classification is found . The image retrieval algorithm consists of two phases : first ,

finding the central image which has the best matching in all the central images, and

second , finding the perfect matching image from the classification which has the

central image . The classification and finding the central image are off-line work , so the

algorithm saves the time of the image retrieval .

According to simulation results，the algorithm is effective.

Another efficient algorithm which based on wavelet for face recognition is also put

forward in this paper. The train images are being transformed by the Daubechies

Wavelet and find the max variance of them . The cross correlation among the wavelet is

being used as the distance to cluster , and the central face image of per classification is

found . The face recognition algorithm is similar with the first image retrieval . The

most important thing is that the transformation of Daubechies Wavelet reduces the

dimension of the image . So the computation is decreased . And the face recognition is

accelerated.

The simulation results show that these algorithms are effective for face

classification and recognition.

Keywords: Face recognition Multi-resolution analysis Entropy

Image recognition Clustering Image classification

Wavelet

中文摘要

ABSTRACT

第一章绪论.....................................................................................................................1

§1.1 研究背景............................................................................................................1

§1.2 研究意义............................................................................................................2

§1.3 研究框架............................................................................................................3

第二章人脸识别研究现状.............................................................................................4

§ 2.1 人脸识别的内涵...............................................................................................4

§2.1.1 人脸检测................................................................................................4

§2.1.2 预处理....................................................................................................4

§2.1.3 特征提取................................................................................................5

§2.1.4 分类识别................................................................................................6

§ 2.2 人脸识别和分类的研究...................................................................................6

§2.2.1 发展历史................................................................................................6

§2.2.2 国内研究现状........................................................................................9

§ 2.3 人脸识别的公共资源.......................................................................................9

§ 2.4 人脸识别研究的热点和难点.........................................................................10

§ 2.5 小结.................................................................................................................10

第三章小波变换的相关理论.......................................................................................11

§3.1 小波变换的发展与历史................................................................................11

§3.2 小波变换及其数学定义................................................................................11

§3.2.1 离散小波变换的定性描述..................................................................13

§3.2.2 小波滤波器..........................................................................................14

§3.2.3 尺度滤波器..........................................................................................15

§3.3 小波变换实质及优点....................................................................................16

§3.4 二维小波变换与图像处理............................................................................17

§3.5 小波基的选取................................................................................................20

§3.6 基于小波的人脸特征的提取........................................................................20

第四章相关图像分类和检索算法研究.......................................................................23

§ 4.1 相关系数判别准则.........................................................................................23

§ 4.2 分类器的设计.................................................................................................24

§ 4.3 图像分类和图像检索.....................................................................................24

§ 4.4 小结.................................................................................................................26

第五章多分辨率和分类的人脸识别算法研究...........................................................31

§ 5.1 分类器的设计.................................................................................................31

§ 5.2 多分辨率分析.................................................................................................32

§ 5.3 人脸识别的算法及实验结果分析.................................................................33

§ 5.4 小结.................................................................................................................35

第六章总结...................................................................................................................36

§ 6.1 小结.................................................................................................................36

§ 6.2 展望.................................................................................................................36

参考文献.........................................................................................................................37

在读期间公开发表的论文和承担科研项目及取得成果.............................................41

致谢.............................................................................................................................42

文档加载中……请稍候！
如果长时间未打开，您也可以点击刷新试试。

下载文档到电脑，查找使用更方便

15 积分 4人已下载

立即下载 VIP免费下载

摘要：

摘要随着计算机互联网技术的发展，如何快速地对大量的人脸图像进行识别，成为近四十多年来的研究热点。因此，人脸识别技术获得了飞速的发展，并在国家反恐、信息安全、访问控制等领域有着广泛的应用。但是研究人员发现，人脸识别仅仅在约束条件下能够达到满意的效果，而在非约束条件下的人脸识别还远未成熟。在人脸识别研究中，如何快速对获得的人脸图像进行识别，并且降低外界条件对于人脸识别的影响，是人脸识别的研究热点之一。本文首先广泛研究了人脸识别的发展历史和研究现状，对人脸识别系统的组成结构、公共资源以及当前的研究热点和难点进行了深入研究。最终将研究重点集中于提高人脸识别速度方面。为提高人脸识别的速度，论文提出了一种...

展开>> 收起<<

多分辨率分析的人脸识别.pdf

共45页,预览5页

还剩页未读，继续阅读

多分辨率分析的人脸识别

相关推荐

开通VIP享超值会员特权

作者详情

相关内容

推荐作者

热门标签

举报选择: