Please wait a minute...
浙江大学学报(工学版)  2017, Vol. 51 Issue (12): 2299-2310    DOI: 10.3785/j.issn.1008-973X.2017.12.001
计算机与通信技术     
采用多重特征蒙板的人像皮肤美化技术
鲁晓卉1, 王进1, 陆国栋1, 张东亮2
1. 浙江大学 流体动力与机电系统国家重点实验室, 浙江 杭州 310027;
2. 浙江大学 国际设计研究院, 浙江 杭州 310027
Portrait skin beautification technology using multiple feature masks
LU Xiao-hui1, WANG Jin1, LU Guo-dong1, ZHANG Dong-liang2
1. State Key Laboratory of Fluid Power Transmission and Control, Zhejiang University, Hangzhou 310027, China;
2. International Design Institute, Zhejiang University, Hangzhou 310027, China
 全文: PDF(23024 KB)   HTML
摘要:

为了解决现有人像皮肤美化技术中出现的细节模糊以及特征丢失问题,提出一种基于构建多重特征蒙板的人像皮肤美化技术.第一重蒙板为利用高反差保留和强光迭加构建的面部显著瑕疵特征蒙板,基于该蒙板可预先去除人像面部的显著瑕疵;第二重蒙板为利用肤色检测构建的人像背景特征蒙板;第三重蒙板为利用Face++算法构建的面部关键区域特征蒙板;第二、三重蒙板的结合使用可以解决人像的背景特征及面部关键区域特征由于后续滤波处理而变模糊的问题;第四重蒙板为利用高反差保留构建的面部立体视觉控制蒙板,该蒙板可在之前的皮肤美化处理基础上还原人像的面部立体视觉特征.实验结果表明,该技术使用较小的皮肤美化系数即可去除瑕疵,并能在皮肤美化系数较大的情况下保留人像关键区域以及立体细节等真实感信息特征,从而解决人像皮肤美化过程中细节清晰度与瑕疵去除程度无法兼得的矛盾,最终达到人像美化的目的.

Abstract:

Portrait skin beautification technology using multiple feature masks was proposed to solve detail blurring and character losing problem in existing portrait skin beautification technologies. The first mask was called facial apparent blemish feature mask. This mask was made up of high pass and superposition of hard light, which was used for removing apparent facial blemishes beforehand. The second mask called portrait background feature mask was made up of skin color detection. The third mask called facial key feature mask was made up of Face++ algorithm. The second mask along with the third mask can help to solve the problem that filtering may blur facial key features and portrait background features. The fourth mask called stereo vision control mask was made up of high pass. This mask can restore facial stereo vision features after portrait skin beautification. Experimental results show that this method could wipe off facial blemishes by setting a smaller portrait skin beautification coefficient and retain realistic features, such as facial key features, as well as facial stereoscopic effect with a larger portrait skin beautification coefficient. Therefore, the contradictory problem that portrait skin beautification should either keep the clarity of details or keep the degree of removing facial blemishes was solved. This technology can achieve the purpose of facial beautification.

收稿日期: 2016-10-27 出版日期: 2017-11-22
CLC:  TP391.41  
基金资助:

国家自然科学基金资助项目(51275460,61472355).

通讯作者: 王进,男,副教授.orcid.org/0000-0003-3106-021X.     E-mail: dwjcom@zju.edu.cn
作者简介: 鲁晓卉(1993-),女,博士生,从事CAD/CG研究.orcid.org/0000-0003-1303-9900.E-mail:luxh@zju.edu.cn
服务  
把本文推荐给朋友
加入引用管理器
E-mail Alert
作者相关文章  

引用本文:

鲁晓卉, 王进, 陆国栋, 张东亮. 采用多重特征蒙板的人像皮肤美化技术[J]. 浙江大学学报(工学版), 2017, 51(12): 2299-2310.

LU Xiao-hui, WANG Jin, LU Guo-dong, ZHANG Dong-liang. Portrait skin beautification technology using multiple feature masks. JOURNAL OF ZHEJIANG UNIVERSITY (ENGINEERING SCIENCE), 2017, 51(12): 2299-2310.

链接本文:

http://www.zjujournals.com/eng/CN/10.3785/j.issn.1008-973X.2017.12.001        http://www.zjujournals.com/eng/CN/Y2017/V51/I12/2299

[1] CHEN D, REN S, WEI Y, et al. Joint cascade face detection and alignment[C]//ECCV. Zurich:Springer, 2014:109-122.
[2] ZHANG K, ZHANG Z, LI Z, et al. Joint face detection and alignment using multi-task cascaded convolutional networks[J]. IEEE Signal Processing Letters, 2016,23(10):1499-1503.
[3] RANJAN R, PATEL V M, CHELLAPPA R. HyperFace:a deep multi-task learning framework for face detection, landmark localization, pose estimation, and gender recognition[J]. IEEE Transactions On Pattern Analysis and Machine Intelligence, 2016(3):1-10.
[4] ZHOU E, FAN H, CAO Z, et al. Extensive facial landmark localization with coarse-to-fine convolutional network cascade[C]//IEEE International Conference on Computer Vision Workshops. Moscow:IEEE, 2013:386-391.
[5] LIN P H, CHEN B H, CHENG F C, et al. A morphological mean filter for impulse noise removal[J]. Journal of Display Technology, 2016, 12(4):344-350.
[6] KULKARNI R K, LAHOTI C B, MEHER S. Impulse denoising using improved progressive switching median filter[C]//ICWET'10 International Conference and Workshop on Emerging Trends in Technology. New York:ACM, 2010:586-590.
[7] BAEK J, JACOBS D E. Accelerating spatially varying Gaussian filters[J]. Acm Transactions on Graphics, 2010, 29(6):81-95.
[8] ADAMS A, GELFAND N, DOLSON J, et al. Gaussian KD-trees for fast high-dimensional filtering[J]. ACM Transactions on Graphics, 2009, 28(3):1-12.
[9] CARLOS A B M. Filtering the shadows from poorlyilluminated photos[J]. ACM Symposium on Applied Computing, 2010:1599-1600.
[10] FUJITA S, FUKUSHIMA N, KIMURA M, et al. Randomized redundant DCT:efficient denoising byusing random subsampling of DCT patches[C]//SIGGRAPH Asia 2015 Technical Briefs. New York:ACM, 2015:7.
[11] LAPARRA V, GUTIERREZ J, CAMPS V G, et al. Image denoising with kernels based on natural image relations[J]. Journal of Machine Learning Research, 2010, 11(1):873-903.
[12] DABOV K, FOI A, KATKOVNIK V, et al. Image denoising by sparse 3D transform-domain collaborative filtering[J]. IEEE Transactions on Image Processing, 2007, 16(8):2080-2095.
[13] SUGIMOTO K, KAMATA S I. Compressive bilateral filtering[J]. IEEE Transactions on Image Processing A Publication of the IEEE Signal Processing Society, 2015, 24(11):3357-3369.
[14] LEE C, SCHRAMM M T, BOUTIN M, et al. Analgorithm for automatic skin smoothing in digitalportraits[J]. IEEE International Conference on Image Processing Pages, 2009:3149-3152.
[15] LIANG L, JIN L, LI X. Facial skin beautificationusing adaptive region-aware masks[J]. IEEE Transactions on Cybernetics, 2014, 44(12):2600-2612.
[16] BRAND M, PLETSCHER P. A conditional random field for automatic photo editing[C]//IEEE Conference on Computer Vision and Pattern Recognition.Ancholage:IEEE, 2008:1-7.
[17] YANG Y, ZHAO H, YOU L, et al. Semantic portrait color transfer with internet images[J]. Multimedia Tools and Applications, 2017, 76(1):523-541.
[18] LIAO Q, JIN X, ZENG W. Enhancing the symmetry and proportion of 3D face geometry[J]. IEEE Transactions on Visualization and Computer Graphics, 2012,18(10):1704-1716.
[19] GUO D, SIM T. Digital face makeup by example[C]//IEEE Conference on Computer Vision and Pattern Recognition. Miami:IEEE, 2009:73-79.
[20] GUO G. Digital anti-aging in face images[C]//IEEE International Conference on Computer Vision. Barcelona:IEEE, 2011:6-13.
[21] GONZALEZ R C, WOODS R E. Digital image processing[M]. 3rd ed. New Jersey:Prentice Hall, 2002:305-308.
[22] DADGOSTAR F, SARRAFZADEH A. An adaptive real-time skin detector based on Hue thresholding:a comparison on two motion tracking methods[J]. Pattern Recognition Letters, 2006, 27(12):1342-1352.
[23] MARKUS W. Caltech:frontal face dataset[EB/OL].[1999-01-01]. http://www.vision.caltech.edu/html-files/.
[24] FEI. Face database[EB/OL].[2006-01-01]. http://fei.edu.br/~cet/facedatabase.html.
[25] KAE A, SOHN K, LEE H, et al. Augmenting CRFs with Boltzmann machine shape priors for image labeling[C]//IEEE Conference on Computer Vision and Pattern Recognition. Portland:IEEE, 2013:2019-2026.

[1] 郑洲, 张学昌, 郑四鸣, 施岳定. 基于区域增长与统一化水平集的CT肝脏图像分割[J]. 浙江大学学报(工学版), 2018, 52(12): 2382-2396.
[2] 蔡晋辉, 张光新, 才辉. 基于连通掩模的重构开算子及应用[J]. J4, 2010, 44(4): 675-680.
[3] 王宣银, 梁冬泰. 基于多元图像分析的表面缺陷检测算法[J]. J4, 2010, 44(3): 448-452.
[4] 孙志海, 孔万增, 朱善安. 视频目标定位的减法聚类改进算法[J]. J4, 2010, 44(3): 458-462.
[5] 毛锋, 张树有, 黄长林. 图像散布图和小波多分辨分析的模具异物检测[J]. J4, 2009, 43(10): 1749-1756.
[6] 范翔, 夏顺仁. 基于特征的显微图像全自动拼接[J]. J4, 2009, 43(7): 1182-1186.
[7] 陈成, 庄越挺, 肖俊. 相机运动条件下的视频前景提取[J]. J4, 2009, 43(6): 975-977.
[8] 张冬梅, 刘利刚. 基于角度滤波的平面图形光顺算法[J]. J4, 2009, 43(6): 1042-1046.