Efficient parallel implementation of the Hough transform on a distributed memory system

摘要

In parallel implementations of the Hough transform it is commonly suggested that either the transform space or the set of image features can be distributed among the processing elements. This paper proposes a method to link parallel access to feature points in the image, and parallel access to the transform space. A synchronous processing sequence is suggested so that both can be distributed. Realtime results have been achieved on a multiple instruction, multiple data (MIMD) distributed image memory and distributed transform memory architecture.