Host: The Institute of Image Electronics Engineers of Japan
This paper estimates detection errors of image rotation angles in our RPSA method in which the frequency spectrum for each image is carefully considered to detect the image rotation angle reliably and accurately. Spatial filtering effects on detection errors are obtained for four types of synthetic images. It is made clear that the image binarization lowers the detection accuracy and smooth-filtering heightens the accuracy, and that errors changes periodically with some frequencies as the rotation angle increases. Detection errors are analyzed from the results and causes of the errors are made clear.