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

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 rectangular objects of known size in an image. A 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.