Image registration (IR) is the process of geometric overlaying or alignment of two or more 2D/3D images of the same scene (unimodal registration), taken or not at different time slots, from different angles, and/or by different image acquisition systems (multimodal registration). Technically, image registration implies a complex optimization of different parameters, performed at local or/and global level. Local optimization methods often fail because functions of the involved metrics with respect to transformation parameters are generally nonconvex and irregular, and global methods are required, at least at the beginning of the procedure. This paper presents a new evolutionary and bio-inspired robust approach for IR, Bacterial Foraging Optimization Algorithm (BFOA), which is adapted for PET-CT multimodal and magnetic resonance image rigid registration. Results of optimizing the normalized mutual information and normalized cross correlation similarity metrics validated the efficacy and precision of the proposed method by using a freely available medical image database. Keywords: medical imaging, image registration, soft computing, evolutionary strategies, bacterial foraging algorithm, global optimization.
About Multimodal Image RegistrationImage registration (IR) is a fundamental task in computer vision used to find either a spatial transformation (e.g., rotation, translation, etc.) or a correspondence (matching of similar image entities) among two (or more) images taken under different conditions (at different times, using different sensors, from different viewpoints, or a combination of them), with the aim of overlaying such images into a common one [1], [2], [3], [4]. Over the years, IR has been applied to a broad range of situations from remote sensing to medical images or artificial vision and CAD systems, and different techniques have been independently studied resulting in a large body of research.IR methods can be classified in two groups according to the nature of images: pixel/voxelbased IR methods (also called intensity-based), where the whole image is considered for the registration process; and, on the other side, feature-based methods, which consider prominent information extracted from the images, being a reduced subset of them. The latter methods take advantage of the lesser amount of information managed in order to overcome the problems found in the former when the images present some inconsistences to deal with, for example,