Image registration in presence of multiplicative noise by Particle Swarm Optimization

Image registration is a really challenging task especially when multiplicative noise is present in the images. The presence of multiplicative noise degrades the image content severely. Derivative based methods like gradient descent and Levenberg Marquardt Algorithm fails to perform image registration in such cases. In this paper we propose a Particle Swarm Optimization (PSO) based image registration technique which is well capable to perform image registration between multiple images that are corrupted with multiplicative noise. The specialty of this method is that its accuracy is high. It is faster and it successfully works when the number of registration parameter is higher. The supremacy of this technique compared to other differentiation based techniques is that, it assures robustness to both the additive and multiplicative noise. It is an iterative approach of optimization where the mean square error between one pair of noisy images is reduced. It deals with four different registration parameters namely the shift along horizontal and vertical direction, rotation and scaling. It can equally perform for large scale as well as small scale disparities between pair of images.