An Improved Particle Swarm Optimization With Dynamically Changing Inertia Weight And Its Application In Medical Image Registration

The union of multi-modal images can apply the information got from different formation equipment to provide more comprehensive information for clinical diagnosis and therapy. The process of image registration is essentially equal to a complex multi-parameter optimization problem. The target of this paper is to search an algorithm not only has a high registration precision but also converges as fast as possible.This paper reviews the signification and concept of registration and compares common registration methods. It also introduces the basic principle of Particle Swarm Optimization (PSO). PSO is a global evolutionary approach which can effectively avoid the local extremum in biomedical image registration. Its strongpoint and weakness are illuminated. Then several improved PSO algorithms based on inertia weight are presented. Their improved effect and shortcoming are illustrated by the result of image registration experiment.A new PSO with dynamically changing inertia weight is proposed. The evolution speed factor and aggregation degree factor is introduced and the weight is formulated as a function of these two factors according to their impact on the search performance of the swarm. A great lot of experiment is performed for the selection of inertia of these two factors. The initial location is assigned uniformly to avoid local extremum caused by evolution in some little area initiated randomly. A parameter h is added to make the iteration cease when it is approximately equal to 1, improving the speed. Experimental result indicates that the new algorithm can have a high registration precision and rapid registration speed.