Info iconThis preview shows pages 1–2. Sign up to view the full content.

View Full Document Right Arrow Icon
FACIAL AGING SIMULATION BASED ON SUPER-RESOLUTION IN TENSOR SPACE Fangyuan Jiang, Yunhong Wang School of Computer Science and Engineering, Beihang University, China ABSTRACT The facial change caused by aging progression might signif- icantly degrade the performance of a face recognition sys- tem. One major way to deal with this problem is to predict the aging process. The work presented in this paper pro- posed a framework to simulate the face aging process by means of super-resolution. Considered the nature of multi- modalities in face image set, multi-linear algebra is intro- duced into the super-resolution method to represent and process the whole image set in tensor space. The simulating results represented in the paper are compared with the ground truth face image of the same people. Index Terms— Aging simulation, super resolution, ten- sor 1. INTRODUCTION The research of facial aging simulation is an interesting and important task since the technique will enhance the robust- ness of a face recognition system upon the aging variation. However, not much work has been devoted to the task since its difficulty in the complex and uncontrolled nature of the aging progress itself. Previous related attempts include average faces and ca- ricature algorithm to develop aging simulation method by Burt in [1], extraction of age change components using PCA and 3D face shape model by C. Choi in [2], statistical shape-intensity combined face model to build aging func- tion by A. Lantis in [3], Hussein [4] discussed the facial deformation based on face anthropometry theory and simu- lated wrinkles with BRDF quotient image technique. Rama- nathan [5] proposed a method for face verification across age based on a Bayesian Classifier. J.Suo [6] proposed a multi-resolution dynamic model for aging simulation task. In this paper, we propose a framework to simulate ag- ing process by means of super-resolution technique. In this framework, a low-resolution face image is super-resolved to a high-resolution face image which keeps the identity in- formation invariant but contain rich age-related feature (mainly in facial textures). During the procedure, the multi- linear view has been introduced to provide an integral sight to the face image set with different modalities. Instead of super-resolving the face image pixel by pixel, we apply su- per-resolution to the identity parameter vector obtained by projecting the input low-resolution image to the tensor basis space constructed from the low-resolution training image set. Specifically, we compute the maximum a posteriori proba- bility (MAP) estimation of the identity parameter vector of the true high-resolution images, and reconstruct the high- resolution images with different ages. This paper is organized as follow. In Section 2, our framework of aging simulation is described. Section 3 in- troduces the tensor algebra. Section 4 describes the super- resolution method we adopted. Experimental results are shown in Section 5 and followed by conclusion in Section 6.
Background image of page 1

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
Image of page 2
This is the end of the preview. Sign up to access the rest of the document.

This note was uploaded on 04/22/2010 for the course MI IP taught by Professor Vladbalan during the Spring '10 term at Universidad del Rosario.

Page1 / 4


This preview shows document pages 1 - 2. Sign up to view the full document.

View Full Document Right Arrow Icon
Ask a homework question - tutors are online