Edge detection using parallel ant colony optimization with Hadoop MapReduce: implementation and scalability

Abstract

The Ant Colony Optimization (ACO) is a popular optimization algorithm that finds use in multiple application areas. Though not among the common uses of this algorithm, edge detection in image analysis is a very functional application of this meta-heuristic. To improve the edge detection capabilities, the inherent parallel nature of the ACO method can be combined with the distributed computing framework provided by the Hadoop/Map-Reduce infrastructure. The latter provides a simple, scalable and fault-tolerant distributed processing paradigm that has been popular in industry and the academic community. In this thesis, we explore the Elastic MapReduce service provided by Amazon Web Services to implement ACO algorithm for edge detection in images, and study its scalability and effectiveness by standard metrics. In addition, we demonstrate a filtering technique to reduce the noisy background of images to achieve significant improvement in the accuracy of edge detection.