DocumentCode :
590689
Title :
Facial image prediction using exemplar-based algorithm and non-negative matrix factorization
Author :
Chang, Hsuan T. ; Peng, H.W.
Author_Institution :
Dept. of Electr. Eng., Nat. Yunlin Univ. of Sci. & Technol., Douliou, Taiwan
fYear :
2012
fDate :
3-6 Dec. 2012
Firstpage :
1
Lastpage :
4
Abstract :
Human aging face prediction is a popular research topic because of its various useful applications such as security system, missing persons search system, etc. In this study, we propose Exemplar-based Algorithm whose property considers the environment of human growth. Moreover, both the non-negative matrix factorization and linear interpolation methods are used to perform the prediction for six facial ROIs. In the proposed method, we employ the family images, in which each family member has more than one images at different ages. And we predict the image ROIs to replace the original ones to obtain the prediction result. However, it is difficult to collect the facial image ROI of families at various age, we also refer the databases from the internet. In experimental results, the correlation coefficient between the real and predicted images can reach 0.82. However, the factor such as expression and light in the reference images could result in lower correlation coefficient.
Keywords :
age issues; correlation methods; face recognition; gesture recognition; interpolation; matrix decomposition; Internet; exemplar-based algorithm; expression factor; facial ROI; facial image prediction; human aging face prediction; human growth environment; light factor; linear interpolation methods; lower correlation coefficient; missing persons search system; nonnegative matrix factorization; regions-of-interact; security system; Aging; Databases; Face; Face recognition; Humans; Image sequences; Interpolation;
fLanguage :
English
Publisher :
ieee
Conference_Titel :
Signal & Information Processing Association Annual Summit and Conference (APSIPA ASC), 2012 Asia-Pacific
Conference_Location :
Hollywood, CA
Print_ISBN :
978-1-4673-4863-8
Type :
conf
Filename :
6411836
Link To Document :
بازگشت