Distributed stereo-correlation algorithm

This paper addresses the problem of parallelizing one of the most computationally intensive imaging tasks, namely, the stereo-correlation operation. Stereo-correlation is a statistical procedure utilized to automatically derive the depth information from a pair of digitized pictures taken from the same scene but at different positions. Thus, the operation automatically generates the third dimension (z-coordinate) of each pixel of the image of the scene. The paper presents a simple but novel distributed stereo-correlation algorithm which uses a data farming approach to balance the work load. It presents the Parallel Virtual Machine implementations of the algorithm together with a set of execution times.