My research efforts are focused on the development of efficient (in terms of both accuracy and performance) computational models and algorithms. The problems or applications that I deal with tend to involve a very large number of numerical computations, and are currently mostly relevant to the fields of image processing and computer vision; I am also interested in simulation and visualization problems.
Since the purpose of developing efficient models is to practically solve a particular class of problems, I do not consider a model and its implementation as separate entities. Rather, knowledge of the context which a model or algorithm is to be implemented in should be incorporated in its design and analysis. In particular, I am interested in exposing parallelism to make better use of modern architectures, particularly those of a data-parallel paradigm, such as GPUs. Oftentimes, a well-known algorithm or computational model cannot fit this context and a new one must be developed.
Presentations
Alexandros-Stavros Iliopoulos, Jun Hu, Nikos Pitsianis, Xiaobai Sun, Mike Gehm, and David Brady. "De-ghosting for Gigapixel Snapshot Processing," in GPU Technology Conference, March 2013, San Jose, CA, USA. (pdf)