In Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy

Ha Le

will defend his dissertation

llumination-invariant Face Recognition

Abstract

Recent advances in face-recognition technology have created many smart products, such as smartphones with face unlock and smart doorbell cameras. However, face-recognition algorithms remain challenging when identifying facial images with many changes in appearance due to poor lighting conditions. The goal of this dissertation is to achieve statistically significant improvement with the performance of face recognition systems using 2D facial images that exhibit variations in illumination. Three primary objectives are proposed, namely (i) collect and annotate facial data from different lighting conditions, (ii) develop and evaluate an algorithm for matching face images to overcome appearance changes due to illumination variations, and (iii) develop and evaluate an algorithm for matching face images to overcome perceptual change due to extreme illumination conditions. This proposal presented two face datasets that facilitate the impact evaluation of face recognition algorithms on illumination variations. Besides, a list of methods on illumination-invariant face recognition is proposed that significantly improves the accuracy over previous work. These methods enhanced the verification performance of face recognition algorithms on poor lighting images by introducing a face relighting framework as a data augmentation method. Finally, the proposal presented a set of methods to tackle the face recognition challenge under low light conditions. The works introduced in this dissertation are capable of successfully identifying human faces in insufficient lighting conditions. Extensive experimental evaluation on the two collected datasets confirmed the robustness of the proposed methods to illumination variations.

Date: Monday, July 13, 2020
Time: 1:00 - 3:00 PM
Place: Online Presentation - Zoom meeting
https://uofh.zoom.us/j/94421726628?pwd=TDY0NmkzNnZ3T0F2S3JROWVDRHAxdz0
Advisor: Dr. Ioannis Kakadiaris

Faculty, students, and the general public are invited.