An automatic image registration method using downhill simplex method and its applications

Abstract

In this paper, we proposed an effective method for time series CT image registration using Mutual Information (MI) and downhill simplex method. By using this method it was possible to obtain correct results in a shorter time. When we include all range of histogram in MI evaluation, the value of MI becomes worse. In order to prevent this problem, we find a suitable range of image histogram in MI calculation by using discriminant analysis. Moreover, in order to speed up the convergence, we propose an effective initial position specification and geometry based finish condition (vector terminate). In the experiment, 3D rigid body alignment is performed successfully by six variables of movement and rotation.