Investigating the Use of Actors for Computer Vision Applications

Abstract

In this paper, we outline an investigation of actor model features by means of the realization of programs solving significant problems. Our aim is to define some initial ideas about a parallel object/oriented programming framework for computer vision, in which all relative aspects can be addressed. Therefore, in the paper we describe an initial experiment of writing parallel code for solving a complex computer vision problem: the segmentation of cursive script characters. Through this task we want to show how an application programmer can express the different kinds of parallelism present in the problem, emphasizing the advantages obtained by the use of the actor model, both from the concurrency and the object-oriented design points of view. The programs have been written by means of a simple but powerful notation which we introduced with the aim of deeply investigating and testing actor model characteristics and as a core for the realization of an actor based high-level language.