Date of Award

7-7-2020

Publication Type

Master Thesis

Degree Name

M.A.Sc.

Department

Electrical and Computer Engineering

First Advisor

Behnam Shahrrava

Second Advisor

Balakumar Balasingam

Keywords

Eye Tracking, Eye-Gaze Estimation, Image Processing, Object Detection, Tracking Object

Rights

info:eu-repo/semantics/openAccess

Creative Commons License

Creative Commons Attribution-Noncommercial-No Derivative Works 4.0 License
This work is licensed under a Creative Commons Attribution-Noncommercial-No Derivative Works 4.0 License.

Abstract

Detecting an extremely small object in a image has always been an important problem. The problem of detecting an object with circular Point Spread Function (PSF) in a focal plane array (FPA) obtained by imaging sensors has several engineering applications. In a recent work, the maximum likelihood (ML) detector was derived for image observations that were corrupted by Gaussian noise in each pixel. The proposed ML detector is optimal under the assumption that the FPA contains a circular object that has its signal intensity spread in multiple image pixels in the form of a Gaussian point spread function (PSF) with known standard deviation. The efficiency of estimation is validated by comparing it with Cramér Rao Lower Bound (CRLB). In this thesis, we develop an approach to estimate the PSF's covariance, noise covariance, and total energy of the signal. In this thesis, we generalize these results to a generic (elliptical) PSF. We applied the proposed method on a real-world application, eye tracking. Eye tracking is emerging as an attractive method of human computer interaction. In the last project included in this thesis, we consider the problem of eye gaze detection based on embedded cameras such as webcams. Unlike infrared cameras, the performance of a conventional camera suffers due to fluctuations in ambient light. We developed a novel approach to improve performance. Further, we implemented our proposed ML approach to detect the center of the iris and showed it to be superior to existing approaches. Using these approaches, we demonstrate an eye gaze estimation approach using the embedded webcam of a laptop.

Share

COinS