Giulio Destri

Decomposition of arbitrarily shaped binary morphological structuring elements using genetic algorithms

A number of different algorithms have been described in the literature for the decomposition of both convex binary morphological structuring elements and a specific subset of nonconvex ones. Nevertheless, up to now no deterministic solutions have been found to the problem of decomposing arbitrarily shaped structuring elements. This work presents a new stochastic approach based on genetic algorithms, in which no constraints are imposed on the shape of the initial structuring element nor assumptions are made on the elementary factors, which are selected within a given set.

Robot self-localization by means of vision

We present an application of vision-based object recognition capabilities to the self-positioning-problem of an autonomous robot. Alphanumeric signs are placed in the robot environment as position markers and perceived through an on-board CCD camera on a pan-tilt head. Sign recognition is performed by a neural network based system, driven by some a-priori knowledge about the characteristics of the objects used as markers (signs). When given a map of the location of markers, the robot is able to estimate its position from the information extracted through perceived images. Marker distances and angular displacements allow the computation of a position uncertainty region for the mobile robot. Even using common, human readable markers, localization is performed with an average position accuracy within a few centimeters.

CELLULAR AUTOMATA AS A COMPUTATIONAL MODEL FOR LOW-LEVEL VISION

In this paper we discuss the use of the Cellular Automata (CA) computational model in computer vision applications on massively parallel architectures. Motivations and guidelines of this approach to low-level vision in the frame of the PROMETHEUS project are discussed. The hard real-time requirement of actual application can be only satisfied using an ad hoc VLSI massively parallel architecture (PAPRICA). The hardware solutions and the specific algorithms can be efficiently verified and tested only using, as a simulator, a general purpose machine with a parent architecture (CM-2). An example of application related to feature extraction is discussed.

Cellular neural networks as a general massively parallel computational paradigm

In this paper is presented the use of the discrete-time cellular neural network (DTCNN) paradigm to develop algorithms devised for general-purpose massively parallel processing (MPP) systems. This paradigm is defined in discrete N-dimensional spaces (lattices) and is characterized by the locality of the direct information transmission between the space points (cells) and by continuous values of data and parameters; the DTCNN paradigm is thus able to express most of the typical MPP applications. A general version of a DTCNN has been implemented and optimized for three MPP architectures, namely the Connection Machines CM-2 and CM-5 and the Cray T3D. The comparison between the three machine performances with those achieved by a standard SPARC-20 workstation shows that, particularly with large lattices, the speed-up allowed in the computational times is significant and the range of solvable problem sizes is widely extended.

Toward the Optimal Decomposition of Arbitrarily Shaped Structuring Elements by Means of a Genetic Approach

The decomposition of a binary morphological structuring element is a well-known problem that has often been addresses in the literature. This work present a new approach based on an Evolution Program: using an iterative stochastic technique, it allows to determine the optimal decomposition of an arbitrarily shaped binary morphological structuring element into the shortest chain of elementary factors chosen from a given set.

Using PVM to develop a distributed object-oriented language for heterogeneous processing

Abstract This paper presents an object-oriented distributed language, called PVOOL, that is a good tool for developing real distributed applications. PVOOL is based on active objects called pv_units and concurrent message passing among them. The main feature of PVOOL is the possibility to coordinate the activity of different pv_units through virtual circuits. Virtual circuits are objects that indicate what are the operations to be performed, what are the pv_units that must do them, and what is the order in which they must be done to give a certain task. Virtual circuits can be dynamically created by pv_units and so this distributed object-oriented system can change its behavior during its life. PVOOL takes advantage of PVM procedures to perform objects distribution and message exchange, without explicitly writing low-level code. The use of PVM allows an easily distribution of a program, that is of its objects, on a heterogeneous network of parallel and serial computers. In particular, the paper describes how the features offered by PVOOL were useful to implement a flexible environment for the development of computer vision applications.

Experiments on the Decomposition of Arbitrarily Shaped Binary Morphological Structuring Elements

The decomposition of binary structuring elements is a key problem in morphological image processing. So far only the decomposition of convex structuring elements and of specific subsets of non-convex ones have been proposed in the literature. This work presents the results of a new approach, based on a Genetic Algorithm, in which no constraints are imposed on the shape of the initial structuring element, nor assumptions are made on the elementary factors, which are chosen from a given set.

Parallel Computation of the Matrix of the Chemical Distances for Defective Graphs

Abstract We report for an efficient parallel SIMD algorithm, implemented on the Connection Machine, calculating the distance matrix on a large class of defective graphs (graphs with vacancies) representing a given chemical structure. The crucial algorithmic aspects are described in details. The first application of our method simulates the diffusion of vacancies in a periodic square lattice under the effects of a novel, pure topological potential: the Wiener number of the graph.

Low-Level computer vision algorithms: Performance evaluation on parallel and distributed architectures

Publisher Summary This chapter discusses the performance evaluation of low-level computer vision (CV) algorithms operating in parallel and distributed environments. In low-level CV algorithms, most of the operations are local the new value associated with each pixel depends only on the values coming from a well-defined and limited neighborhood of that pixel. Therefore, low-level CV problems are suited to be ported on a parallel or distributed environment as they show also the most balanced behavior from the point of view of the computation versus communication ratio. The cellular neural network (CNN) paradigm is appropriate to describe this kind of computation as it embodies , as special cases, all CV problems solved with algorithms involving local operations. Hence, the use of CNNs to evaluate the performance of CV applications in parallel environments is the best choice as CNNs are both a superset of all local low-level CV algorithms and suited for parallelization. A measurement taken with respect to the most general CNN formulation can give an effective value of the lower performance bound offered by a given platform for CV applications.

MPP Solution of Lattice Problems: Data Parallel vs. Message Passing

This paper compares the performance achieved by a Data Parallel and a Message Passing implementations of the Cellular Neural Network computational paradigm on two parallel architectures, CM-5 and T3D. CNN embodies a wide set of lattice problems, characterized by the locality of information exchanges among lattice points.