摘要近年来，人脸识别技术得到了飞速地发展。然而进行人脸识别时，品质差的图像会严重地影响人脸识别系统的准确性。所以，对图像进行品质评估是提高人脸识别准确率的一个关键问题 。为了解决这个问题，本文对图像品质评估算法进行了研究。首先，介绍了当前主要的图像品质评估算法；其次，在此基础上,本文提出一种新的人脸图像品质评估算法,主要评估了人脸的尺寸、姿态、光照对称性、对比度以及亮度。结合主观评价经验设置权重，并计算得出一个总得分。最后，使用人脸检测程序包检测人脸的68个显著点，获得显著点坐标，然后在Matlab环境下进行实验。实验结果表明，实验获得的机器评分与人的主观评分很相似。33109
关键词 人脸图像 显著点 图像品质 客观评价
毕业论文设计说明书外文摘要
Title Face image quality assessment algorithm based landmarks
Abstract
In recent years, the technology of face recognition has been developed rapidly. However, when the face recognition, the quality of the image will seriously affect the accuracy of the face recognition system. Therefore, the quality assessment of the image is a key problem to improve the accuracy of face recognition. In order to solve this problem, the image quality assessment algorithm is studied. Firstly, this paper introduces the main image quality assessment algorithm; Secondly, based on these algorithms , this paper puts forward a new face image quality assessment algorithm, which mainly assesses the face size, pose, illumination symmetry, contrast, and brightness.weights are set to calculate a total score combined with the subjective evaluation experience .Finally, 68 landmarks of the face were detected by face detection program, and the coordinates of the human face landmarks were obtained, and the experiments were carried out under the Matlab environment.Experimental results show that the machine score obtained by the experiment is similar to the subjective score.
Keywords face image landmark quality objective evaluation
目 次
1 引言 1
1.1 背景 1
1.2 国内外研究现状 1
1.3 图像品质评估方法的展望 2
1.4 本文结构 2
2 人脸品质评估概述 3
2.1 主观评价和客观评价 3
2.1.1 主观评价 3
2.1.2 客观评价 3
2.2 影响客观评价的常见因素 4
2.2.1 光照 4
2.2.2 表情和姿态 4
2.2.3 遮挡问题 5
2.2.4 人脸的大小 5
2.3 OpenCV 及 Matlab 概述 6
2.3.1 OpenCV 6
2.3.2 Matlab 6
3 人脸品质评估算法设计 7
3.1 人脸品质评估算法概述 7
3.1.1 人脸尺寸评估算法 7
3.1.2 人脸位置评估算法 7
3.1.3 人脸倾斜度评估算法 7
3.1.4 图像对比度评估算法 8
3.1.5 图像亮度评估算法 8
3.2 改进后的图像品质评估算法 9
3.2.1 人脸尺寸评估算法 9
3.2.2 基于显著点的人脸姿态评估算法 9
3.2.3 人脸光照对称性评估 11 基于显著点的人脸图像品质评估算法:http://www.751com.cn/jisuanji/lunwen_30011.html