NONLINEAR OPERATORS FOR IMAGE PROCESSING: DESIGN, IMPLEMENTATION AND MODELING TECHNIQUES FOR POWER ESTIMATION

BERNACCHIA, GIUSEPPE

NONLINEAR OPERATORS FOR IMAGE PROCESSING: DESIGN, IMPLEMENTATION AND MODELING TECHNIQUES FOR POWER ESTIMATION

BERNACCHIA, GIUSEPPE

2000-01-25

http://thesis2.sba.units.it/store/handle/item/12679

http://hdl.handle.net/10077/11391

Doctoral Thesis

Contributor(s)

LONGO, O. GIUSEPPE

Abstract

In the past few years, multimedia application have been growing very fast, being applied to a large variety of fields. Applications like video conference, medical diagnostic, mobile phones, military applications require to handle large amount of data at high rate. Images as well as voice data processing are therefore very important and they have been subjected to a lot of efforts in order to find always faster and effective algorithms. Among image processing algorithms, we believe that rational operators assume an important role, due to their versatility and effectiveness in data processing. In the last years, several algorithms have been proposed, demonstrating that these operators can be very suitable in different applications with very good results. The aim of this work is to implement some of these algorithm and, therefore, demonstrate that rational filters, in particular, can be implemented without requiring large sized systems and they can operate at very high frequencies. Once the basic building block of a rational based system has been implemented, it can be successfully reused in many other applications. From the designer point of view, it is important to have a general framework, which makes it able to study various configurations of the system to be implemented and analyse the trade-off among the design variables. In particular, to meet the need far versatile tools far power estimation, we developed a new macro modelling technique, which allows the designer to estimate the power dissipated by a circuit quickly and accurately. The thesis is organized as follows: In chapter 1 we present some of the algorithms which have been studied for implementation. Only a brief overview is given, leaving to the interested reader some references in literature. In chapter 2 we discuss the basic architectures used for the implementations. Pipelined structures have been mainly used for this thesis, but an overview of the nowaday available approaches for timing optimization is presented. In chapter 3 we present two of the implementation designed for this thesis. The approaches followed are ASIC driven and FPGA drive. They require different techniques and different solution for the design of the system, therefore it is interesting to see what can be done in both the cases. Finally, in chapter 4, we describe our macro modelling techniques for power estimation, giving a brief overview of the up to now proposed techniques and showing the advantages our method brings to the design.