Nonlinear Correlation Measurement Based Registration for Sequential Images

Registration for sequential images has significant meanings in clinical diagnosis and analysis. Many researches are concentrated on using higher-order mutual information (HMI) to fulfill the registration task. But HMI has some disadvantages, which limit the applications of HMI to a large degree. The paper proposes a new registration measure, called nonlinear correlation information entropy (NCIE), which is revised from mutual information (MI) and can be easily sensitive to the amount of nonlinear correlation information between multiple variables. Using this measure we conduct some simulations. The registration results testify that the registration method based on maximization of NCIE is a very useful and efficient method for sequential images registration.