A New Non Linear Inertia Weight Approach in PSO for Faster Rigid Image Registration
Sanjeev Saxena, Mausumi Pohit · 2019
A non-linear inertia weight approach is proposed and validated for faster image registration. Inertia weight in Particle Swarm Algorithm controls the movement of the particles and effectively balances the exploitation and exploration mechanism of the swarm. The most efficient approach used in literature is that of linear inertia weight variation during the PSO run. In the present work a Non-linear approach is used for image registration problem. For the registration mutual information is used as a similarity metric. Images are registered using the standard linearly varied inertia weight approach and the new non-linear approach and comparative study is done. The results show that the proposed inertia weight variation is conducive in faster convergence of PSO and hence provide faster image registration. Further, to validate our study the proposed approach is used on twenty standard CEC benchmark functions and the comparison is done with the standard approach. It was found that the convergence rate was improved with the proposed inertia weight variation.