Title:	Hallucinating Very Low Resolution Face Images to 16x magnification with Age based Attributes
Authors:	Kim, Jihwan Choi, Sunghee
Citation:	Journal of WSCG. 2020, vol. 28, no. 1-2, p. 57-63
Issue Date:	2020
Publisher:	Václav Skala - UNION Agency
Document type:	článek article
URI:	http://wscg.zcu.cz/WSCG2020/2020-J_WSCG-1-2.pdf http://hdl.handle.net/11025/38425
ISSN:	1213-6972 (print) 1213-6980 (CD-ROM) 1213-6964 (on-line)
Keywords:	halucinace obličeje;potrubní síť;věk;osobnost;hluboké učení
Keywords in different language:	face hallucination;pipeline network;age;personality;deep learning
Abstract in different language:	Face hallucination is a type of super resolution that restores very low resolution (8 8 pixel) to high resolution (128 128 pixel) face images. Since unique facial features caused by age, e.g.wrinkles, are ignored during restoration, restored face images can be somewhat dissimilar to the original faces, particularly for older people. To solve this problem, we construct a pipeline network to restore face images more realistically by including age attribute, predicted from the low resolution image. Predicted age attribute is divided into young and old groups, where the aging network is the last pipeline stage and only applied when the original face image includes old age attributes. Thus, older people tend to be restored with wrinkles and features similar to their original appearance. Restored images are compared qualitatively and quantitatively with images created by existing methods. We show that the proposed method maintains and restores age related personality features, such as wrinkles, producing higher structural similarity index than other methods.
Rights:	© Václav Skala - UNION Agency
Appears in Collections:	Volume 28, Number 1-2 (2020)

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.