Hough transform for transputer arrays

This paper focuses on the real-time implementation of a line detection algorithm based on the Hough transform using standard transputer hardware. The main topics discussed are concerned with: programming the transputer array for real-time image processing applications; and loading and flushing of the image data into and out of the array. The study is based on a comparison of two parallel programming models of spatial parallelism: the control- and the demand-driven model. In the control-driven model, the controller sends the image to the processors in the array which perform the desired task exchanging information when necessary. At the end of this computation, the results are sent back to the controller. In the demand-driven model there is a valve processor and many worker processors. The image is subdivided into small partitions and the workers make requests to the valve for the next partition of the image