Abstract
Attempts have been made to speed up image-understanding computation involving conventional serial algorithms by decomposing these algorithms into portions that can be computed in parallel. Because many classes of algorithms do not readily decompose, one seeks some other basis for parallelism (i.e, for using additional hardware to obtain higher processing speed). In this paper we argue that “parallel guessing” for image analysis is a useful approach, and that several recent IU algorithms are based on this concept. Problems suitable for this approach have the characteristic that either “distance” from a true solution, or the correctness of a guess, can be readily checked. We review image-analysis algorithms having a parallel guessing or randomness flavor. We envision a parallel set of computers, each of which carries out a computation on a data set using some random or guessing process, and communicate the “goodness” of its results to its co-workers through a “blackboard” mechanisms.
Share this
