Chapter 11

Metaheuristics for Continuous Variables. The Registration of Retinal Angiogram Images

11.1. Introduction

Image registration is an important tool that is used to resolve any problems that may arise during the analysis of medical images. In the past, scientists have tried to apply traditional minimization strategies to the field of image registration. Such traditional minimization strategies include exhaustive search, gradient descent algorithm, simplex optimization method, simulated annealing algorithm, genetic algorithms and Powell’s search [RIT 99, JEN 01].

In most cases image registration is carried out in two phases: image processing and then optimization of a similarity criterion. The aim of the image processing phase is to improve the quality of the image and to extract the relevant information which can be used to improve the optimization phase. The aim of the optimization phase is to find optimal modifications of the image, in accordance with a particular objective function which describes the quality of the image registration. The optimization phase is often carried out by using specific optimization methods. The methods used tend to be local searches, that are unable to find any global optimum.

Since the image processing phase and the calculation of the objective function requires a lot of time, global optimization methods such as metaheuristics (which require a high number of evaluations) are avoided and local optimization methods are used instead since they ...

Get Optimisation in Signal and Image Processing now with the O’Reilly learning platform.

O’Reilly members experience live online training, plus books, videos, and digital content from nearly 200 publishers.