Campus Event Calendar
Event Entry
What and Who
'Learning Based Prediction of Children's Facial Growth'
Kristina Scherbaum
IMPRS
Master's Seminar
AG 1, AG 2, AG 3, AG 4, AG 5, SWS
MPI Audience
English
Note: We use this to send email in the morning.
Date, Time and Location
Wednesday, 24 May 2006
13:05
60 Minutes
E1 4
024
Saarbrücken
Abstract
We present a complete framework for the simulation of facial growth in
images of babies or children, based on a 3D Morphable Model. Our approach involves
methods for data acquisition, example-based simulation of growth, and
image processing. From a group of 84 babies and 241 teenagers, we recorded several scans per
person, as well as photographs from different viewing directions of the teenagers. We
propose a novel, model-based algorithm for processing the scans, which
includes 3D registration, stitching, hole filling, and establishing correspondence
to the existing Morphable Model. A homogeneous database is developed although the input datasets differ in many fundamental properties. In our acquisition pipeline a colouring technique was applied to the greyscale textures of the baby scans. For the teenager scans additional high-resolution textures are captured from photographs of the teenagers. To simulate aging, a piecewise linear approach shifts the faces along the directions that connect the averages of different age groups. For aging simulation in images, a 3D model of the initial face is reconstructed and after applying an age-transformation on shape and texture, the face is rendered back into images of children at appropriate target ages. Our approach has a variety of potential applications, such as helping to find missing children.
images of babies or children, based on a 3D Morphable Model. Our approach involves
methods for data acquisition, example-based simulation of growth, and
image processing. From a group of 84 babies and 241 teenagers, we recorded several scans per
person, as well as photographs from different viewing directions of the teenagers. We
propose a novel, model-based algorithm for processing the scans, which
includes 3D registration, stitching, hole filling, and establishing correspondence
to the existing Morphable Model. A homogeneous database is developed although the input datasets differ in many fundamental properties. In our acquisition pipeline a colouring technique was applied to the greyscale textures of the baby scans. For the teenager scans additional high-resolution textures are captured from photographs of the teenagers. To simulate aging, a piecewise linear approach shifts the faces along the directions that connect the averages of different age groups. For aging simulation in images, a 3D model of the initial face is reconstructed and after applying an age-transformation on shape and texture, the face is rendered back into images of children at appropriate target ages. Our approach has a variety of potential applications, such as helping to find missing children.
Contact
Kerstin Kathy Meyer-Ross
226
--email hidden
passcode not visible
logged in users only
Jennifer Gerling, 05/19/2006 10:30 -- Created document.