A Harris-based Region Detector on a Computational Grid

This paper introduces a new Harris-based feature detector algorithm, characterized by no parameters tuning to detect region of interest. Preliminary results show that the proposed methodology returns good results with respect to standard detectors which need a set of parameters. An uncommon parallel implementation of the proposed algorithm is presented to support the high computational workload which is required to avoid the approximation of the results. Our parallel approach differs from the conventional one because an internal scheduler, based on the expected running time, is used to balance the data distribution on a client-server model. The aim of this paper is also to underline the advantages offered by the parallel architecture to implement image feature detectors, as a starting point to new image analysis applications