Supervisor: Prof. Dr. Cevdet Aykanat

An iterative image restoration method is formulated by converting the restoration problem into a linear feasibility problem. For the solution of the linear feasibility problem, surrogate constraint methods are utilized which works efficiently for large size problems and are amenable for parallel implementations. Among the proposed methods, the basic method and an improved version of the parallel method is utilized. Using several partitioning strategies and adopting different communication models, several parallel implementations are performed. The implementations are evaluated based on the per iteration performance and on the overall performance. Also, the effect of the partitioning on the convergence rate is investigated.

keywords: Parallel image restoration, distortion, parallel algorithms, linear feasibility, surrogate constraint method, hyper graph partitioning, rowwise partitioning, checkerboard partitioning, fine-grain partitioning, point-to-point communication, all-to-all communication, convergence rate.

DATE: January 20, 2003, Tuesday @ 10:00
PLACE: EA-502