Facial detection and recognition resilient to physical image deformations

hh.sePublications

Simple search

Advanced search -
Research publications Advanced search -
Student theses Statistics

Change search

ExportLink to record

Permanent link

Direct link

https://hh.diva-portal.org/smash/project.jsf?pid=project:242

BETA

Project

Project type/Form of grant

Project grant

Title [sv]

Ansiktsdetektering och robust igenkänning med avseende på bild deformationer

Title [en]

Facial detection and recognition resilient to physical image deformations

Abstract [sv]

Detta projekt fokuserar på mänskliga ansikten under ogynnsamma verkliga förvärvs/registreringsförhållanden. Det omfattar detektering av ansiktets landmärken såväl som personigenkänning med statistiska lokala mått(features) som extraheras från ansiktsbilder. Det är relaterat till metoder för tillförlitlig detektering och igenkänning av ansikten i förhållanden som innefattar varierande skala, upplösning, huvudrotation och ansiktsocklusion.

Abstract [en]

The project is focused on facial detection and recognition resilient to real-world alterations such as changing scale, head rotation or occlusion. They are important tasks in many applications from human computer interaction to biometrics and forensics. Although these tasks are easily solved by humans, current Computer Vision algorithms struggle in the presence of such alterations.One goal is to contribute with methods to reliably detect facial landmarks with symmetry filters, which have been successfully applied to detect eyes in controlled conditions. This contrasts with existing algorithms, where face is modeled holistically and detected globally, thus failing in case of perturbations. For recognition, we will use Gabor filters on retinotopic sampling grids. Such a grid models the photoreceptors in the retina, with sampling frequency decreasing exponentially from the center to the periphery. Gabor filters mimic the functioning of cells of the visual cortex having different spatial directions and frequencies. This approach has already shown high discrimination power using face in controlled environments.These innovations will contribute to increase user convenience and comfort, reducing the cooperation level required and allowing the use of their own device and natural interaction patterns to communicate with digital systems. It will enable the use of facial data acquired in a wide range of operational conditions due to the large amount of cameras already available.

Principal InvestigatorAlonso-Fernandez, Fernando Halmstad University, School of Information Technology, Halmstad Embedded and Intelligent Systems Research (EIS), CAISR - Center for Applied Intelligent Systems Research (Closed down 2021-12-17)
Co-InvestigatorBigun, Josef Halmstad University, School of Information Technology, Halmstad Embedded and Intelligent Systems Research (EIS), CAISR - Center for Applied Intelligent Systems Research (Closed down 2021-12-17)

Coordinating organisation

Halmstad University

Funder

Vetenskapsrådet

Period

2013-01-01 － 2016-12-31

Keywords [sv]

ansiktsdetektering, ansiktsigenkänning, biometri

Keywords [en]

face detection, face recognition, biometrics, structure tensor

National Category

Human Computer InteractionSignal Processing

Identifiers

DiVA, id: project:242Project, id: 2012-04313_VR

Search in DiVA

On the subject

Human Computer Interaction Signal Processing

Search outside of DiVA

Google Google Scholar

v. 2.47.0

WCAG

Halmstad University Library

Info for students

Info for researchers