A comparative image analysis of discrete radial Fourier transforms

Fourier transform is not often applied to pattern and object recognition, because of its inefficiency in image description and calculation. In order to solve this problem, a bi-discrete radial cosine transform is proposed. This new transform has two superior properties. Firstly, by using Mukundan’s square-to-circular transformation, it projects images in square to images in discrete circle girds. Secondly, different from the classical radial orthogonal plane, the new orthogonal plane is composed of discrete cosine functions orthogonal in both radial and circumferential directions. Experiment results indicate that the proposed bi-discrete radial cosine transform has an excellent performance in terms of efficient image reconstruction and computation.