Normalization of rotationally symmetric shapes for pattern recognition

摘要

Image normalization is useful for image recognition by the matching method. Recently, a normalization algorithm was developed to normalize images under the distortion of translation, rotation, scaling and skew. However, the method fails when the image is rotationally symmetric. A method is introduced called the modified Fourier descriptor to normalize rotationally symmetric shapes. This method is simple and fast, and can be combined with the preceding normalization algorithm to make it more general and realizable.