An efficient 3D gradient-based algorithm for medical image registration using correlation-coefficient maximization

Most registration algorithms for medical images may suffer from slow convergence and sensitivity to initialization. In this paper, we propose an efficient gradient-based algorithm for rigid intensity-based registration of medical images. It takes advantage of gradient-descend approach for maximizing the correlation coefficient similarity measure. Furthermore, we automatically adjust the optimization rate of algorithm for weak local minima avoidance and convergence acceleration. Experimental results demonstrated superior performance of the proposed algorithm compared to Powell's method (as a non-gradient algorithm) in terms of CPU time and solution quality for two sets of medical images.