Three-dimensional cement image registration based on Multi-layer PSO and mutual information

For the research of cement hydration, it is necessary to carry out the three-dimensional (3D) image registration of cement in order to obtain the dynamic change process of the 3D cement microstructure. The optimization of similarity measure is needed to match the intensity of registration. Due to the complex functions of these parameters, it is difficult to find the global optimal solution for local optimization techniques, which requires global optimization methods. The searching layers of the swarm are increased from two layers to multiple layers in the Multi-layer particle swarm optimization (MLPSO). The MLPSO improves the performance of the traditional particle swarm algorithm by this way. In this paper, it uses mutual information(MI) as the similarity measure of 3D image registration and uses the MLPSO to solve the spatial transformation parameters. The experiment indicates that the method shows a better result and improves the convergence speed.