Autonomy Oriented Computing Applied to Image Processing and Feature Extraction

This paper presents an overview of Autonomy Oriented Computing (AOC) and related approaches to image processing and feature extraction problems. The basic elements of such AOC systems are autonomous entities placed in an environment. The environment, in our case, is viewed as a twolayer 2D lattice containing an image in the 1 layer and a notice board at the 2 layer. The environment serves as the place where autonomous entities reside, roam and operate. The goal of each autonomous entity, here viewed as a distributed computational agent, is to effectively locate and label image feature pixels, by exhibiting a number of reactive and rational behaviors, such as diffusion, breeding and communication. The entity behavioral repository related to interaction with the image, consists of operations such as testing whether certain pixel color intensity falls into a specific range of values or if the observed pixel match with the specified criteria of the regions homogeneity. It also have operations such as a convolution with the specified mask or evaluation of the array of the pixel values which entity has memorized during it’s search through the image. Each autonomous entity while selecting which behavior to execute is directed by its state, behavior functions, and/or interaction with other entities. Similar techniques have been performed with related optimization algorithms based on artificial ant colonies (Ant Colony Optimization (ACO)) and Particle Swarm Optimization (PSO). Autonomy Oriented Computing can be seen as a more general form of the PSO and ACO algorithms, where agent entities are defined generally with behavioral traits not only directed to optimization, but to any parallel application which can be performed through the interaction of many loosely coupled entities.