edoc

Detailed face ageing with morphable models

Schneider, Andreas. Detailed face ageing with morphable models. 2019, Doctoral Thesis, University of Basel, Faculty of Science.

[img]
Preview
PDF
282Mb

Official URL: http://edoc.unibas.ch/diss/DissB_13260

Downloads: Statistics Overview

Abstract

This thesis is about the automatic, photorealistic, and detailed ageing of faces in images. Morphable Models of the face provide a distribution over 3D face shape and texture and are often used for this task. However, they only model features that are at the same location between individuals, which is not the case of wrinkles, spots, pores, and other high-frequency details. We propose a face ageing model capable of producing detailed age progressed images showing global shape and skin tone changes, as well as details and self-shadowing effects. Our face ageing approach consists of different parts that we combine into a single ageing simulation. 1) Ageing in Morphable Model space providing the coarse shape deformation and skin tone change associated with age. 2) An efficient global illumination model making renderings more realistic with self-shadowing. 3) Detail synthesis by transferring the high-frequency parts from a selected example face scan. 4) A wrinkle propagation model to make the transfer specific to wrinkle growth. 5) A stochastic pigment spot model to introduce the spots known to appear with age. We evaluate parts of the approach with a perception experiment showing that the resulting images are perceived as older and similarly realistic as real ones. With the resulting model, we age faces in portrait images and achieve photorealistic results showing detailed face ageing.
Advisors:Vetter, Thomas and Blanz, Volker
Faculties and Departments:05 Faculty of Science > Departement Mathematik und Informatik > Informatik > Computergraphik Bilderkennung (Vetter)
UniBasel Contributors:Schneider, Andreas and Vetter, Thomas
Item Type:Thesis
Thesis Subtype:Doctoral Thesis
Thesis no:13260
Thesis status:Complete
Bibsysno:Link to catalogue
Number of Pages:1 Online-Ressource (viii, 106 Seiten)
Language:English
Identification Number:
Last Modified:11 Sep 2019 04:30
Deposited On:10 Sep 2019 08:23

Repository Staff Only: item control page