Fusion of Color and Depth Information for Facial Recognition using a Multi Perspective Approach | [Proceedings of the Pakistan Academy of Sciences: A. Physical and Computational Sciences • 2017]

Author(s):

1. Mudassir Aman: Electrical Engineering Department, University of Engineering and Technology,Peshawar,Pakistan

2. Waqar Shah: Electrical Engineering Department, University of Engineering and Technology,Peshawar,Pakistan

3. Ihtesham-ul-Islam: Department of Computer Science and IT, Sarhad University,Peshawar,Pakistan

4. Bilal Jan: Department of Computer Science and IT, Sarhad University,Peshawar,Pakistan

5. Muhammad Amir: Electrical Engineering Department, University of Engineering and Technology,Peshawar,Pakistan

6. Bilal-ur-Rehman: Electrical Engineering Department, University of Engineering and Technology,Peshawar,Pakistan

Abstract:

Facial-recognition is an explored and demanding task. Previously, mostly color (RGB) images were used to tackle it. Recently, advances in 3D scanners have been providing extra facial information. This new information improves the performance of current facial recognition architectures. In this research, both RGB and depth image information were utilized for addressing the problem of facial-recognition and by characterizing each image with the use of multi-perspective-approach (MPA). Data were combined from different textural-image-descriptors (TIDs) while keeping the most relevant features. Feature vectors resulting from such combinations were entered into a random-forest-classifier (RFC) to obtain a comparative analysis through the EURECOM facial dataset. The outcomes of our case-studies are comprehensively elaborated in this paper.

Page(s): 365-374

DOI: DOI not available

Published: Journal: Proceedings of the Pakistan Academy of Sciences: A. Physical and Computational Sciences, Volume: 54, Issue: 4, Year: 2017

Keywords:

face recognition , RGBD source , 3D face , Feature Fusion , 2D face , depth image

References:

References are not available for this document.

Citations

Citations are not available for this document.

Citations

Downloads

Views