Contribution à la conception des filtres bidimensionnels non récursifs en utilisant les techniques de l’intelligence artificielle : application au traitement d’images

Abstract

The design of finite impulse response (FIR) filters can be formulated as a non-linear optimization problem reputed to be difficult to solve by conventional approaches. The constraints are high and a large number of parameters have to be estimated, especially when dealing with 2D FIR filters. In order to improve the performance of conventional approaches, we explore several evolutionary algorithms capable of handling large spaces. We propose a new genetic algorithm in which some innovative concepts are introduced to improve the convergence and make its use easier for practitioners. The key point of our approach stems from the capacity of the genetic algorithm (GA) to adapt the genetic operators during the genetic life while remaining simple and easy to implement. Most of parameters and operators are changed by the GA itself. Then, the Particle Swarm Optimization (PSO) is proposed for FIR filter design. The study underlines the importance of introducing intelligence in metaheuristics to make them more efficient by inserting self tuning strategies. Finally, a hybrid genetic algorithm (HGA) is proposed for the design of digital filters. The algorithm is composed of a pure genetic process and a dedicated local approach. The pure genetic process is devoted to the exploratory part of the HGA. The local approach concerns convergence of the algorithm and is highly optimized so as to be tractable. The main contribution is the development of a strategy for guiding the algorithm, in which the parameters auto-adapt by themselves according to the evolution of research solutions. We demonstrate on various cases of filter design a significant improvement in performance. Our contribution mainly concerns the usability of GAs for the specific area of FIR filter design. It seeks to address the current challenge of democratizing the use of GAs for real world problems. Experiments performed with various types of filters highlight the recurrent contribution of hybridization in improving performance. The experiments also reveal the advantages of our proposal compared to more conventional filter design approaches and some reference GAs in this field of application.