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

High dynamic range imaging for face matching

Tools
- Tools
+ Tools

Suma, Rossella (2020) High dynamic range imaging for face matching. PhD thesis, University of Warwick.

[img]
Preview
PDF
WRAP_Theses_Suma_2020.pdf - Submitted Version - Requires a PDF viewer.

Download (23Mb) | Preview
Official URL: http://webcat.warwick.ac.uk/record=b3714355~S15

Request Changes to record.

Abstract

Human facial recognition in the context of surveillance, forensics and photo-ID verification is a task for which accuracy is critical. In most cases, this involves unfamiliar face recognition whereby the observer has had very short or no exposure at all to the faces being identified. In such cases, recognition performance is very poor: changes in appearance, limitations in the overall quality of images - illumination in particular - reduces individuals’ ability in taking decisions regarding a person’s identity.

High Dynamic Range (HDR) imaging permits handling of real-world lighting with higher accuracy than the traditional low (or standard) dynamic range (LDR) imaging. The intrinsic benefits provided by HDR make it the ideal candidate to verify whether this technology can improve individuals’ performance in face matching, especially in challenging lighting conditions. This thesis compares HDR imaging against LDR imaging in an unfamiliar face matching task. A radiometrically calibrated HDR face dataset with five different lighting conditions is created. Subsequently, this dataset is used in controlled experiments to measure performance (i.e. reaction times and accuracy) of human participants when identifying faces in HDR.

Experiment 1: HDRvsLDR (N = 39) compared participants’ performance when using HDR vs LDR stimuli created using the two full pipelines. The findings from this experiment suggest that HDR (µ =90.08%) can significantly (p< 0.01) improve face matching accuracy over LDR (µ =83.38%) and significantly (p<0.05) reduce reaction times (HDR 3.06s and LDR 3.31s).

Experiment 2: Backwards-Compatibility HDR (N = 39) compared par ticipants’ performance when the LDR pipeline is upgraded by adding HDR imaging in the capture or in the display stage. The results show that adopt xi ing HDR imaging in the capture stage, even if the stimuli are subsequently tone-mapped and displayed on an LDR screen, allows higher accuracy (capture stage: µ =85.11% and display stage: µ =80.70%), (p<0.01) and faster reaction times (capture stage: µ =3.06s and display stage: µ =3.25s), (p< 0.05) than when native LDR images are retargeted to be displayed on an HDR display.

In Experiment 3: the data collected from previous experiments was used to perform further analysis (N = 78) on all stages of the HDR pipeline simultaneously. The results show that the adoption of the full-HDR pipeline as opposed to a backwards-compatible one is advisable if the best values of accuracy are to be achieved (5.84% increase compared to the second best outcome, p<0.01).

This work demonstrates scope for improvement in the accuracy of face matching tasks by realistic image reproduction and delivery through the adoption of HDR imaging techniques.

Item Type: Thesis (PhD)
Subjects: T Technology > TA Engineering (General). Civil engineering (General)
T Technology > TR Photography
Library of Congress Subject Headings (LCSH): Human face recognition (Computer science), High dynamic range imaging
Official Date: May 2020
Dates:
DateEvent
May 2020UNSPECIFIED
Institution: University of Warwick
Theses Department: Warwick Manufacturing Group
Thesis Type: PhD
Publication Status: Unpublished
Supervisor(s)/Advisor: Chalmers, Alan
Sponsors: Seventh Framework Programme (European Commission) ; Warwick Manufacturing Group ; Rabin Ezra Scholarship Trust
Format of File: pdf
Extent: xv, 162 leaves : illustrations (some colour)
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