Parallel implementation of the Hough transform for the extraction of rectangular objects

L. Hopwood, W. Miller, A. George
Proceedings of SOUTHEASTCON '96  
In image processing applications, the storage capacity required for images can exceed feasible storage capabilities. A technique to alleviate this problem by removal of unnecessary background information through image processing is discussed. Specifically, a parallel implementation of a first-order, derivative-based edge detection algorithm and the Hough transform applied to rectangular objects is given. A variation of the classical Hough transform to detect lines is employed to locate
more » ... ar objects of known size in an image. Parallel Virtual Machine is used to exploit the inherent parallelism found in these algorithms over a cluster of 7 workstations. Through the use of these techniques, the rectangular object is detected and stored as a separate image, and storage capacity can be reduced by approximately 30%, not including standard data compression. Parallelizing the algorithms provides a significant speedup advantage over the normal sequential operation of the programs.
doi:10.1109/secon.1996.510069 fatcat:hfehtoacrfbsxjquljo2s7t5wq