Parallel algorithms for plane detection using an adaptive Hough transform

摘要

Two variations of an adaptive Hough transform for plane detection have been implemented on a multipleinstruction-multiple data architecture based on transputers. In each case, process and processor allocation is data-dependent. Load balancing is achieved by a mixture of geometric decomposition and the processor farm paradigm. Evaluation of the algorithms on synthetic and real image data show their viability in terms of plane detection and increased speed of parallel computation.