 |
Face De-occlusion |
| By exploiting the power of hallucination to “see things that were not originally there”, our technique provides the first reasonable solution to the face occlusion problem. We remove the occlusion by predicting the appearance of the occluded region using the best un-occluded region through a model-based Bayesian formulation. Our technique is useful as a pre-processing step for other face processing tasks that require high quality face images, e.g. face recognition. |
|
 |
 |
Face Recognition Across Pose and Illumination Using Image Synthesis |
| In this project, we propose to use image synthesis to recognize faces. During the training stage, we synthesize images under different illumination and pose to augment our training set. Given the augmented training set, different classifiers can be explored to do recognition. |
|
|
 |
Exploring Face Space |
| Face space is defined as the set of face images under any possible poses and illuminations. Using computer graphics technique, we synthesize face images under different poses and illuminations. After reducing dimension, we model face space with pdf by using Gaussian Mixture Model. Given one pdf for each person, we analyze these pdfs for where they overlap, where they do not. |
|
|
 |
Continuous Authentication |
| For most computer systems, once the identity of the user has been verified at login, the system resources are typically made available to the user until the user exits the system. This is not appropriate for high-security environments. To reduce this window of vulnerability, we propose a system that continually verifies the presence/participation of a logged-in user. This is done by integrating multimodal passive biometrics (face+fingerprint) in a Bayesian framework that combines both temporal and modality information holistically, rather than sequentially. |
|
|
 |
Face Cloning for Animation |
| A Structure-Driven-Adaptation (SDA) method was proposed to efficiently reconstruct animated 3D faces of real human individuals. The technique is based on adapting a prototype generic facial model to the acquired surface data in an “outside-in” manner: deformation applied to the external skin layer is propagated along with the subsequent transformations to the muscles, with the final effect of morphing the underlying skull. |
|
