Radial Krawtchouk moments for rotational invariant pattern recognition

Krawtchouk moments are not rotational invariant. Hence, this paper introduces a new radial Krawtchouk moments using polar representation of an image by a one dimensional orthogonal discrete weighted Krawtchouk polynomials and a circular function. Further rotational invariants are derived from the proposed radial Krawtchouk moments. In order to test the proposed approach, three problems namely image reconstruction, rotational invariance and texture recognition are attempted using the proposed moments. In order to assess the quality of the reconstructed image, reconstruction error, peak signal to noise ratio (PSNR) and universal image quality index (UIQI) are computed. It is observed from the simulation results that the reconstructed image converges to the original image as the order of the moments is increased. Further more, the difference in rotational invariants before and after the rotation of an image is very small. Finally, the test texture images are correctly recognized from a set of texture images that are available in a small database.