Automatic Image Registration Based Control Points and Evaluation
In this paper, we present a fully automatic image registration method, applied to a pilot study. Image registration plays an important role to the medical image field for the patient prognosis and diagnosis. Particularly image standardization and image correlation could be the most significant studies. Therefore, we have created several feasible two dimensional color training images with the objects having different major axes and sizes. The training images were rotated, translated, scaled and sheared. Consequently, we calculated the registered correction by the minimum squared error(MSE) of the between control points distances and object sizes. Then, we obtained a valuable result (>0.04). It could be useful to biomedical image registration.
Index Terms - Automatic image registration, control points, standardization, color training images, registered correction, minimum squared error