This paper describes a new approach for rapidly and accurately detecting rotation center positions of rotated images and correcting the images.Rotation center position detection and correction of rotated images are important in fields such as LSI and micro-machine productions.In the proposed approach,rotation center position detection and correction of the rotated images can be performed simply and precisely without any repetitive procedure.In the experiments,reference images were synthetically rotated by 0.2∼4.0 degrees with the step of 0.2 degrees to generate input images rotated around many coordinates widely distributed from the center to the boundary in the reference images.The detection accuracy of the rotation centers tends to depend on the true rotation center position and the average of absolute errors of the rotation center detections for all rotation centers is 2.40 pixels.To estimate the correction accuracy,rotation angles and position displacements between the corrected images and the reference images were detected.The correction accuracy for image rotation is about 0.0129 degrees on average and that for the displacement is about 0.052 pixels on average.From the results,it is clear that the proposed approach makes it possible to accurately detect the rotation center position and also to precisely correct the image rotated around an arbitrary position.