I. De Lotto

Image compression by the wavelet decomposition

Decomposition of images with the Haar orthonormal basis which is an important member of compactly supported Wavelets and a quadtree structured hierarchical coding technique are used in this work to obtain high image compression efficiency and time complexity linear in the number of pixels. An exhaustive testing of the algorithm has been performed on images of different complexity and typical of some application environment (image transmission and storing, remote control of intelligent robots). The results of the experiments arc presented and discussed. Finally, a comparison of quality performance of these techniques with the JPEG (Block Cosine Transform coding) compression technique is presented.

A template-matching operator for edge-points detection in digital pictures

Abstract A fast operator that extracts edge-points of objects in digitized images is presented. The technique can be classified as a local template-matching method, however it allows considerable improvement over several known template matching techniques in terms of boundary thickness, sensitivity both to strong and to faint edges, insensitivity to intensity and to smooth variations of scene illumination and, finally, of computational cost.

Monte Carlo Metropolis simulation of interacting anisotropic polarizable spins on a lattice

We present a Monte Carlo Metropolis computer simulation study of the equilibrium configurations of a spin lattice model of anisotropically polarizable particles. In the presence of an external field the spins polarize and couple via dipolar interactions. To simulate the system we used both the complete Hamiltonian, requiring heavy calculations of the local field, and a simplified version, obtained by a first order expansion of the local field. We find that in both cases, the dipolar interaction induces a new and unexpected orientational order characterized by a depressed value of the orientational order parameterhP2i and by a corresponding enhanced value ofhP4i.

Ewald potentials evaluated through look-up tables

Abstract In this paper we present a computational method to calculate the potential of long-range interactions between particles which allows considerable saving in computing time. The method is based on the tabulation of the potential due to particles located on a sufficiently accurate 3D grid. The actual potential is evaluated by a three-linear interpolation of the tabulated values. The validity of the method is proved through a Monte Carlo—Metropolis simulation of ionic fluids. Finally, the speed-up achieved executing the method on different workstations is shown.

A development and simulation environment for a floating point operations FPGA based accelerator

Exploiting hardware devoted to a specific application requires a proper programming support, like libraries allowing a simple interface with the device. As a complementary design, a dedicated device language might be developed to make its programming easier. In previous works by G. Danese et al. (2002) and G. Danese et al. (2003) we presented the architecture of a floating point operations accelerator based on Field Programmable Gate Array (FPGA) technology. In this paper we describe the development environment which allows writing, translating and simulating instruction sequences written in a language specifically conceived and designed for that device.

A parallel neurochip for neural networks implementing the reactive tabu search algorithm: application case studies

In this work we present two different applications implemented on the neurocomputer Totem Nc3001 from Neuricam Inc. The goal of the experimentation is to test, on real problems, the performance of this powerful parallel unit consisting of 32 Digital Signal Processors (DSPs) and to evaluate its suitability to neural network applications. The first problem implemented is a typical classification algorithm in which the network recognises which points belong to different regions inside a 2D space. The second problem is more computationally heavy and consists of a network able to reproduce the eye movements, if properly stimulated. A comparison is reported between Matlab implementations or handwritten code run on workstations and the performance obtained from the Totem chip.

Adaptive restoration of scintigraphic images

Abstract An adaptive technique to restore scintigraphic images is presented. A few examples are discussed to show its performance when the procedure is implemented on a minicomputer with limited core memory.

An image restoration technique: applications to nuclear medicine images

A new procedure for image restoration is presented which involves an estimation of the original image power spectrum to approximate the op timum linear filter. Applications to nuclear medicine images are discussed.

A parallel processor for neural networks

We present two different algorithms implemented through neural networks on a multiprocessor device. The parallel single-chip TI TMS32C80 Multimedia Video Processor (MVP). The goal of this experimentation is to test, on real problems, the performance of this powerful unit made up by one Master Risc Processor and by four Slave Digital Signal Processors (DSPs) and to evaluate its suitability to neural network applications. The first problem implemented is a typical classification algorithm in which the network recognises which points belong to different regions inside a 2D space. The second problem is more computationally heavy and consists of a network able to recognise handwritten digits. The parallel version of the first algorithm, was also tested on a commercially available supercomputer.

A parallel architecture for metropolis simulation of systems with long range interactions

A parallel architecture is presented which is based on the Intel i860 RISC-processor and split into two subgroups of processors: the first designed to execute Monte Carlo loops, and the second to assess the statistical parameters. This approach enables efficient parallelization because data-transfer recurrence among processor groups is relatively small. With regard to the Monte Carlo-Metropolis algorithm (1953) the authors expect to reach performances higher than those given by present generation of CRAYs.<<ETX>>