Research on Image Registration Algorithm Based on Membrane Computing

Image registration is a precondition for image fusion. It is provided with significant research value. For image registration, traditional intelligent algorithm has some shortcomings, such as low registration accuracy and slow convergence.A multi-modal image registration algorithm under the framework of membrane computing is proposed based on the parallel property of membrane computingand it is named as GA-MCIR algorithm. A cell-like P system of membrane structure is designed. Each object in membranes represents a group of transform parameters of floated images. All objects of each elementary membranes are constantly evolved by genetic algorithm, in which the best parameters are obtained and they are transformed into upper-layer membrane. At the same time, the best objects of each elementary membrane reciprocally exchange in the process of evolution. The rules are applied and the upper-layer membranes preserve global optimum objects in current evolution and transform them into sub-membranes to implement the evolution of the included sub-membranes. Finally, the global optimal object is stored in the skin membrane. In order to test the superiority, this study utilizes the multi-modal images, such as, computer tomography(CT)image of the brain and visible and thermal infrared image and is further compared with those images based on GA and PSO respectively. The comparison results reveal that GA-MCIR algorithm obtain better registration accuracy and global convergence.