摘 要

随着生活节奏的加快和皮肤健康问题的不断产生，皮肤护理日益受到人们的重视。但由于缺少便携精确的皮肤检测分析工具和有针对性的建议，人们在挑选护肤品时常感觉无从下手，导致护肤效果不佳。目前国内外的肤质检测仪主要有基于图像分析的大型肤质检测仪和基于生物电阻抗测量法的金属探测式肤质检测仪，前者体积大、价格昂贵，专用于医疗美容行业，难以普及；后者只能检测水分和油分，难以满足消费者的需求。针对用户对肤质检测和保养的个性化需求，本文设计一种基于图像处理的肤质检测算法。

根据皮肤的组织结构，本文提出基于颜色模型、纹理模型和阈值分割模型的肤质检测算法。首先将RGB图像转化为HSV颜色空间，进而对饱和度与亮度进行二值化处理，运用逻辑运算合并处理结果，得出水分、油分、色素的检测结果并进行标记；在此基础上进行形态学运算实现对毛孔的边缘提取，计算毛孔近似圆半径描述毛孔大小；粗糙度的检测采用灰度共生矩阵的纹理分析方法，通过求灰度图共生矩阵4个参数的均值和方差来描述皮肤粗糙度情况。

实验结果表明，本文所设计的肤质检测算法可以实现水分、油分、色素、粗糙度、毛孔五项指标的检测与提取，并且同市面上主流肤质检测仪的检测结果相符，达到预期目标。

关键词：图像处理，肤质检测

Abstract

With the increasingly fast pace of life and continual skin health problems, people have gradually paid attention to skin care. But for lack of portable and precise skin detective tools and specific advice, people often feel confused when choosing skin care produces, which results in bad skin care effects. At present, at home and abroad, there are mainly two kinds of skin detective machine. One is called large skin detector, which is based on visual analysis, the other one is called metal skin detector, which is based on bioelectrical impedance measurement. The former is large in cubage and high in price, so it is hard to be universal and used in cosmetic field. The latter can only detect moisture and oil of skin, which can hardly meet the needs of consumers. To meet the individual needs of consumers in skin detection and skin care, this article proposes a kind of skin detective calculation based on image processing.

According to the composition structure of skin, this article gives a skin image analysis calculation based on color model and threshold segmentation model. Through the establishment of the color model, the percent of water and oil taken up in face skin image, which is considered as a detective indicator, can be further processed to obtain the content of water, oil and pigment. Threshold segmentation and morphological calculations can be used to measure the border of pores and describe the size of pores. Texture analysis which is based on Gray Level Co-occurrence Matrix can be used to distill roughness.

The results of experiments show that the skin detective calculation can detect and distill five indicators including water, oil, pigment, roughness and pores. And the detection results can match that of the main skin detector in market, and reach expected goal.

Keywords: image processing, skin detection

目 录

第1章 绪论 1

1.1 设计背景及意义 1

1.2 国内外研究现状 2

1.3 设计基本内容 4

第2章 肤质检测仪体系结构设计 5

2.1 设计需求与理论基础 5

2.2 设计方案论证 6

2.2.1 颜色空间 6

2.2.2 图像分割 8

2.2.3 纹理分析 10

2.3 本章小结 11

第3章 皮肤图像处理算法的设计与实现 12

3.1 肤质检测算法总体设计 12

3.2 人脸肤质指标提取 13

3.2.1 色素指标提取 13

3.2.2 水分、油分指标提取 17

3.2.3 毛孔指标提取 18

3.2.4 粗糙度指标提取 21

3.3 本章小结 22

第4章 算法调试与结果 23

4.1 算法调试 23

4.2 肤质检测结果 26

4.3 本章总结 30

第5章 总结与展望 31

参考文献 32

致 谢 33

第1章 绪论

1.1 设计背景及意义

随着人们生活节奏的加快，皮肤健康问题伴随压力与焦灼而不断产生，严重影响了人们的精神面貌与身心健康，人们越发的意识到皮肤护理的重要性。对皮肤保养方面存在的误区以及肤质检测与分析方面专业工具和手段的欠缺导致了护肤品的乱用与滥用现象。因此，人们对肤质检测仪的需求与日俱增。

当今国内外的肤质检测仪主要可以分为两类，一类是基于生物电阻抗测量法的金属探测式肤质检测仪，另一类是基于图像处理的大型肤质检测仪。