Resarch On Medical Image Registration Based On The Maximum Of Mutual Information

Medical image registration, which is of great value in research and application, is the core issue of medical image. The key technology is to find out image registration algorithm with high speed, high accuracy and robust matching results. The algorithms based on polynomial registration, correlation functions registration and even recently sequential similarity, neural network all have certain limitations. The image registration method based on the mutual information can remedy those defects.The medical image registration is discussed in this paper based on the maximum of mutual information, combining Powell algorithm, genetic algorithm and particle swarm optimization (PSO) separately. Powell algorithm is a local optimization algorithm, whose computation efficiency is low, so this article mainly introduced the latter two global intelligent optimization algorithms.The first improvement is a precise medical image registration method based on mutual information and genetic algorithm. This method introduces genetic algorithm into registration to improve the search strategy and solves space transformation parameters. Results of experimental show that the algorithm avoids the problem of registration parameter is not global optimal, achieves high registration accuracy and has high speed.The other improvement integrates the advantages of mutual information and particle swarm optimization (PSO) algorithm. An image registration algorithm based on the mutual information and particle swarm optimization is proposed. This algorithm uses uniform assignment to initial position of particle swarm, avoiding initial position concentrates in some small areas and makes the optimal into the local extreme, and also joins evolution speed factor as a search suspend conditions to accelerate the convergence speed.In experimental design and analysis, a lot of comparisons are made among these three algorithms which based on mutual information and Powell algorithm, mutual information and genetic algorithm, mutual information and the particle swarm optimization. Experimental results show that the proposed intelligent optimization algorithm can have good efficiency and stability, which also show the algorithm introduced in this paper has much potential in developments and applications in the future.