FRAME - Facial and Retinal Analysis, Modelling and Estimation

Project description

 The FRAME project aims at 3D reconstruction of biological form, applied to the retina and human faces.

In the case of 3D reconstruction of the retina, the goal is to develop algorithms which are able to distinguish between variations in the healthy eye versus the deformations present in the abnormal eye. The reconstruction will be performed by the use of uncalibrated images. Such an stereoscopic investigation of pathological changes of the retina, will be a complicated task, due to the relative short baseline available.

Face

In the part involving modeling of the human face, 3D scans of form and texture will be used to estimate a 3D Active Appearance Model (AAM). The AAM will be applied to segment new images, both 2D and 3D, which then may be classified. This has a number of important biomedical applications, such as the prediction of human facial growth, and identification of abnormal growth patterns and appearance due to e.g. cranio-facial malformations.

Also a robust classification of human form may be a very desirable task in the future, as an extensive integration of machines is to be expected (eg. robots, computers, transportation, internal house environment etc.). In order to integrate machines at a high level in the society, it is of outmost importance that they are enabled to recognise, eg. human -form, -identity and -expression. And that they are able to do this independent of scale, location and orientation.

Modeling of the shape, texture and dynamic will be done on the basis of a 3D face database, which is under construction in collaboration with the 3D Laboratory at the School of Dentistry, University of Copenhagen.

Implementation and integration of the project will include methods from modern multivariate statistics, computer vision, image analysis, and computer graphics to construct realistic models. These models will be based on decomposition of the variations in the dataset by the use of explorative data analysis. This decomposition is to be done by the use of generalized principal component analysis, principal component regression, independent components analysis, partial least squares, ridge regression and alternatives.

The work in the project will be founded on numerous activities at IMM, DTU, at the Mathematical Imaging Group at LTH, at the 3D laboratory, School of Dentistry, and at the Department of Ophthalmology, Herlev Hospital.

These include retinal image analysis, 2D face models using AAMs, several projects on biological form and projects involving 3D reconstruction from 2D images.

Back to previous page