An Accurate Method of Multi-mode Image Registration Based on Mutual Information and Gradient-weighted
Zuo Xiao-n · Journal of Nanchang University · 2012
This paper proposed an accurate method of multi-mode image registration based on gray-scale mutual information and gradient weighted normalized mutual information as a similarity measuring,and decreasing concave function of the trade-off proportion of the improved particle swarm optimization(PSO) as a search strategy.Registration method based on gray-scale mutual information often considers only grayscale,ignoring or introducing improperly the feature information in image space.As a result,the registration was easy to fall into local minima,and mismatches.The proposed algorithm merged the gradient weighted-to-gray and the mutual information,taking into account the gray statistical correlation of two images and the image spatial characteristics,so as to improve the multi-mode image registration accuracy and stability.Simulation and practical registration on remote sensing images and MRI-PET medical image registration showed the algorithm's good performance:the algorithm was stable and had a high registration precision and parameter accuracy.