The manual analysis of human biometric identifiers in non-pathological retinal fundus images
: An evaluation of presented methodologies and a survey of identifiers within open source databases

  • Lilly Dan

Student thesis: Master's ThesisMaster of Science


From the initial proposal of the use of retinal fundus images to enable human identification in 1935, identity biometrics based upon digital representations of the retinal fundus have been proposed as a highly effective modality applied for high level security applications. As complex internal surface, the retina is highly reflective of ocular and systemic pathological conditions, resulting in a large research base in image analysis of fundus images within the active field of computer- aided diagnostics (CAD).

Subsequently, many researchers in biometric identification utilise the numerous clinical open-source databases to evaluate their proposed image analysis techniques regardless of the variation introduced by the development of ocular pathology. However, a direct assessment of baseline non-pathological human variance within the retina across and within open source databases had not yet been conducted.

To assess the variation of biometric identifiers across datasets, a qualitative assessment of biometric identifiers was conducted; confirming the suitability of superficial retinal vasculature as a foremost identifier for human differentiation whilst highlighting its limitations.

This was followed by the manual analysis of retinal fundus images from a subset of open source databases using an evaluated novel manual analysis protocol in the biological image-analysis program; Fiji. In its current form, the novel manual protocol lacks repeatability over multiple observers and requires revision. The assessment of multiple retinal identifiers was found to be highly impacted by differences in resolution between the databases. A resolution hierarchy was created which confirmed that measurements and the amount of data collected decreased with samplings at lower resolutions.
Date of Award2020
Original languageEnglish
SupervisorLucina Hackman (Supervisor) & Helen Langstaff (Supervisor)


  • Retina
  • Identification
  • Fundus Images
  • Biometrics

Cite this