In this paper, we present a novel descriptor for feature extraction by using a combination of ridgelets and Fourier transform. We have successfully implemented ridgelets on the circular disk containing the pattern and applied Fourier transform on the resulting ridgelet coefficients to extract rotation-invariant features for pattern recognition. The descriptor is very robust to Gaussian noise even when the noise level is high. Experimental results show that the new descriptor is a very good choice for pattern recognition.
展开▼