A Parallel Symbolic Language for Vision Analysis

The authors present an extended version of Paralation Lisp that we developed and used to program, in a unified and user friendly parallel programming environment, a large part of the computer vision algorithms. They describe the computer vision area, motivating the need for the notion of structured parallelism. Next, the Paralation Lisp flexibility, that allows one to cope with structured parallelism, is described. Furthermore, how to build a complex hierarchy of objects relating to vision applications, and how to attach a different form of parallelism to each one is also described starting from the paralation notion. Finally, current research efforts in developing the parallel implementation strategies of the Paralation Lisp on a distributed memory parallel system (DMPS) is described