PhD Dissertation Prospectus Defense: Ahmet Tuysuzoglu

  • Starts: 11:00 am on Monday, October 22, 2012
Robust Inversion and Detection Techniques for Improved Imaging Performance -- Date: Monday, October 22, 2012, 11:00am -- 8 Saint Marys Street, Room 339 -- Chair-Advisor: M. Selim Unlu (ECE) Committee: W. Clem Karl (ECE), David Castanon (ECE), Bennet Goldberg (Physics) --The quality of outcomes of an imaging system depends on two major processing steps, inversion and detection. Inversion refers to the cumulative image formation/reconstruction and restoration efforts involved in an imaging system. Once an image is obtained, it is manually or automatically processed for extraction of any relevant information, which is known as the detection step. The quality of outcomes can be improved by enhancing inversion and detection techniques employed in the imaging process. In this thesis, we take three approaches to improve the performance of imaging systems. We first focus on advanced inversion techniques. As widely reported, conventional inversion techniques are not robust in the face of changing imaging conditions and are prone to exhibit artifacts that hinder imaging performance. We propose a regularized inversion method that incorporates prior information about the underlying field into the inversion framework. We use experimental ultrasound data to show inversion results with the proposed formulation and compare it with conventional inversion techniques. Second, we consider imaging scenarios where it is impractical to use advanced inversion techniques whereas robust detection techniques can be employed to improve imaging outcomes. In particular, we focus on a portable, low cost interferometric imaging platform that is capable of imaging nanoparticles and nano pathogens as small as 40 nm in various experimental scenarios, such as detection of certain proteins and viruses in serum or whole blood. The detection of particles of interest is challenging due to existence of other particles and imaging artifacts. We propose robust detection algorithms using tools from computer vision and principles of nano-optics to efficiently identify particles of interest. We further seek to improve detection performance by utilizing multiple observations of the underlying field, each emphasizing a different set of features. These different observations are simple to collect, only requiring the use of simple optical components hence not compromising the portability and the affordability of the platform. In the final part of the thesis, we aim to develop methods that can combine inversion and detection steps in a unified framework to improve imaging performance. This framework is applicable for cases where the underlying field is label-based such that each pixel of the underlying field can only assume values from a discrete, limited set. We consider this unified framework in the context of combinatorial optimization and propose graph-cut based methods that would result in label-based images, therefore eliminating the need for a separate detection step.
Location:
8 Saint Mary’s Street, Room 339