Rotation invariant feature extraction by combining denoising with Zernike moments

Rotation invariant feature extraction is a classical topic in pattern recognition. It is well known that Zernike moment features are invariant with regard to rotation. However, due to noise present in the unknown pattern image, Zernike moment features can fail to recognize the noisy pattern. In this paper, a new feature extraction method is proposed by combining a wavelet-based denoising method with zernike moment feature extraction in order to achieve improved classification rates. Experimental results demonstrate its superiority over zernike moments without denoising.