当前位置>主页 > 期刊在线 > 信息技术 >

信息技术2018 年3 期

基于稀疏表示与人脸结构的人脸幻构技术研究
黎兆文,胡晓
(广州大学机械与电气工程学院,广东 广州 510006)

摘  要:虽然基于稀疏表示的方法重构人脸图像有着样本库需求小的优点,同时对于平滑区域的恢复也有很好的效果,但是人脸成分以及轮廓边缘细节仍然较为模糊。为了解决这一问题,本文提出了结合稀疏表示的梯度估计边缘优化方法,该方法利用样本库中高分辨率的人脸成分以及边缘梯度统计空间对低分辨率输入人脸进行细节恢复和边缘锐化。实验结果表明,该方法对人脸图像结构的细节恢复有较为理想的表现效果。


关键词:稀疏表示;梯度估计;人脸结构;人脸幻构



中图分类号:TP391.41         文献标识码:A         文章编号:2096-4706(2018)03-0001-05


Face Animation Technology Based on Sparse Representation and Face Structure
LI Zhaowen,HU Xiao
(School of Mechanical and Electrical Engineering,Guangzhou University,Guangzhou 510006,China)

Abstract:Although the reconstruction of face images based on sparse representation has the advantages of small sample base andgood effect on the recovery of the smooth region,the face composition and the outline edge details are still relatively vague.In order tosolve this problem,this paper proposes a gradient estimation edge optimization method combined with sparse representation,which useshigh resolution face components and edge gradient statistical space in sample database to restore and sharpen the edges of low resolutioninput faces.Experimental results show that this method has a satisfactory effect on detail restoration of face image structure.

Keywords:sparse representation;gradient estimation;face structure;face magi


参考文献:

[1] S.Baker,T.Kanade.“Hallucinating faces,”in Fourth IEEEInternational Conference on Automatic Face and Gesture Recognition.Proceedings,IEEE,2000:83-88.

[2] C.Liu,H.-Y.Shum,and C.-S.Zhang.“A two-step approachto hallucinating faces:global parametric model and local nonparametricmodel,”in Proceedings of the 2001 IEEE Computer Society Conferenceon Computer Vision and Pattern Recognition,2001(1):1–192.

[3] W.T.Freeman,T.R.Jones and E.C.Pasztor,“Examplebasedsuper-resolution,” Computer Graphics and Applications,IEEE,2002,22(2):56-65.

[4] YANG J C,WANG Z W,LIN Z,et al.Coupled dictionarylearning for image super-resolution [J].IEEETransactions on ImageProcessing,2012,21(8):3467-3478.

[5] WRIGHT Y J,HUANG TS. Image super-resolution via sparserepresentation [J]. IEEE Transactions on Image Processing,2010,19(11):2861-2873.

[6] ZEYDE R,ELAD M,PROTTER M.On single imagescaleupusing sparse representations [C].//International Conference on Curvesand Surfaces.Springer-Verlag,2010:711-730.

[7] AI N,PENG J Y,ZHU X,et al.SISR via traineddoublesparsity dictionaries [J].Multimedia Tools andApplications,2015,74(6):1997-2007.

[8] W.T.Freeman,E.C.Pasztor,O.T.Carmichael,“Learninglowlevel vision,”Int. J.Comput. Vis,2000,40(1):25-47.

[9] H.Chang,D.-Y.Yeung,Y.Xiong.“Super-resolutionthroughneighbor embedding,” in Proc. IEEE Conf. Comput.Vis.PatternClass,2004(1).

[10] S.Baker and T.Kanade,“Limits on super-resolution andhow to breakthem,”IEEE Trans.Pattern Anal.Mach. Intell,2002,24(9):1167-1183.

[11] Li Y,Liu J,Yang W,et al.Multi-pose face hallucinationvia neighbor embedding for facial components [C].//IEEE InternationalConference on Image Processing,IEEE,2015:4037-4041.

[12] C.-Y.Yang,S.Liu,and M.-H.Yang..“Structured facehallucination,”in Proc.IEEE Int’l Conf.Computer Vision and PatternRecognition,2013.

[13] M.F.Tappen and C.Liu.A Bayesian approach to alignmentbasedimage hallucination. In ECCV,2012.

[14] C.Liu,H.-Y.Shum,and W.T.Freeman.Face hallucination:Theory and practice. IJCV,2007,75(1):115–134.

[15] X.Ma,J.Zhang,and C.Qi.Hallucinating face bypositionpatch. PR,2010,43(6):2224–2236.

[16] R.Fattal.Image upsampling via imposed edge statistics. InSIGGRAPH,2007.

[17] J.Sun,Z.Xu and H.-Y.Shum.Image superresolution usinggradient profile prior. In CVPR,2008

[18] GROSS R,MATTHEWS I,COHN J,et al.Multi-pie[C].2008-06-15,S.l.:s.n.,2008:1-8.

[19] M.Irani and S.Peleg.Improving resolution by imageregistration [J].CVGIP,1991,53(3):231–239.


作者简介:

黎兆文(1991.08-),男,汉族,广东人,硕士。研究方向:人脸图像处理。

通信作者:胡晓(1969-),男,湖南人,教授,硕士。研究方向:智能信号处理;人脸检测和识别;医学信号处理。