Skip to content Skip to navigation
University of Warwick
  • Study
  • |
  • Research
  • |
  • Business
  • |
  • Alumni
  • |
  • News
  • |
  • About

University of Warwick
Publications service & WRAP

Highlight your research

  • WRAP
    • Home
    • Search WRAP
    • Browse by Warwick Author
    • Browse WRAP by Year
    • Browse WRAP by Subject
    • Browse WRAP by Department
    • Browse WRAP by Funder
    • Browse Theses by Department
  • Publications Service
    • Home
    • Search Publications Service
    • Browse by Warwick Author
    • Browse Publications service by Year
    • Browse Publications service by Subject
    • Browse Publications service by Department
    • Browse Publications service by Funder
  • Help & Advice
University of Warwick

The Library

  • Login
  • Admin

Unconstrained face recognition with occlusions

Tools
- Tools
+ Tools

Wei, Xingjie (2014) Unconstrained face recognition with occlusions. PhD thesis, University of Warwick.

[img]
Preview
PDF
WRAP_THESIS_Wei_2014.pdf - Requires a PDF viewer.

Download (23Mb) | Preview
Official URL: http://webcat.warwick.ac.uk/record=b2753542~S1

Request Changes to record.

Abstract

Face recognition is one of the most active research topics in the interdisciplinary areas of biometrics, pattern recognition, computer vision and machine learning. Nowadays, there has been significant progress on automatic face recognition in controlled conditions. However, the performance in unconstrained conditions is still unsatisfactory. Face recognition systems in real-world environments often have to confront uncontrollable and unpredictable conditions such as large changes in illumination, pose, expression and occlusions, which introduce more intra-class variations and degrade the recognition performance. Compared with these factor related problems, the occlusion problem is relatively less studied in the research community.

The overall goal of this thesis is to design robust algorithms for face recognition with occlusions in unconstrained environments. In uncontrollable environments, the occlusion preprocessing and detection are generally very difficult. Compared with previous works, we focus on directly performing recognition with the presence of occlusions. We deal with the occlusion problem in two directions and propose three novel algorithms to handle the occlusions in face images while also considering other factors.

We propose a reconstruction based method structured sparse representation based face recognition when multiple gallery images are available for each subject. We point out that the non-zeros entries in the occlusion coefficient vector also have a cluster structure and propose a structured occlusion dictionary for better modelling them. On the other hand, we propose a local matching based method Dynamic Image-to-Class Warping (DICW) when the number of gallery images per subject is limited. DICW considers the inherent structure of the face and the experimental results confirm that the facial order is critical for recognition. In addition, we further propose a novel method fixations and saccades based classification when only one single gallery image is available for each subject. It is an extension of DICW and can be also applied to deal with other problems in face recognition caused by local deformations.

The proposed algorithms are evaluated on standard face databases with various types occlusions and experimental results confirmed their effectiveness. We also consider several important and practical problems which are less noticed (i.e., coupled factors, occlusions in gallery or/and probe sets and the single sample per person problem) in face recognition and provide solutions to them.

Item Type: Thesis or Dissertation (PhD)
Subjects: T Technology > TA Engineering (General). Civil engineering (General)
Library of Congress Subject Headings (LCSH): Human face recognition (Computer science)
Official Date: August 2014
Dates:
DateEvent
August 2014Submitted
Institution: University of Warwick
Theses Department: Department of Computer Science
Thesis Type: PhD
Publication Status: Unpublished
Supervisor(s)/Advisor: Li, Chang-Tsun
Extent: xiv, 144 leaves : illustrations, charts
Language: eng

Request changes or add full text files to a record

Repository staff actions (login required)

View Item View Item

Downloads

Downloads per month over past year

View more statistics

twitter

Email us: wrap@warwick.ac.uk
Contact Details
About Us