This paper presents a robust, automated registration algorithm, which may be applied to several types of medical images, including CTs, MRIs, X-rays, Ultrasounds and dermatological images. The proposed algorithm is intended for imaging modalities depicting primarily morphology of objects i.e. tumors, bones, cysts and lesions that are characterized by translation, scaling and rotation. An efficient deterministic algorithm is used in order to decouple these effects by transforming images into the log-polar Fourier domain. Then, the correlation coefficient function criterion is employed and the corresponding values of scaling and rotation are detected. Due to the non-linearity of the correlation coefficient function criterion and the heavy computational effort required for its full enumeration, this optimization problem is solved using an efficient simulated annealing algorithm. After the images alignment in scaling and rotation, the simulated annealing algorithm is employed again, in order to detect the remaining values of the horizontal and vertical shifting. The proposed algorithm was tested using different initialization schemes and resulted in fast convergence to the optimal solutions independently of the initial points.