Feature Extraction Using Radon, Wavelet and Fourier Transform

In this paper, we propose a novel descriptor for invariant pattern recognition by using the Radon transform, the wavelet transform, and the Fourier transform. The Radon transform can capture the directional features of the pattern image by projecting the pattern onto different orientation slices. The combination of the 2-D shift invariant wavelet transform with the Fourier transform can extract features that are invariant to rotation of the patterns. Standard normalization techniques are used to normalize the input pattern image so that it is translation and scale invariant. Experiments conducted in this paper show that the proposed descriptor achieves high recognition rates for different combinations of rotation angles and noise levels. The descriptor is very robust to Gaussian white noise even when the noise level is very high.

I. Introduction

Feature extraction is a crucial step in invariant pattern recognition [1]. In general, good features must satisfy the following requirements. First, intraclass variance must be small, which means that features derived from different samples of the same class should be close (e.g., numerically close if numerical features are selected). Secondly, the interclass separation should be large, i.e., features derived from samples of different classes should differ significantly. Furthermore, features should be independent of the size, orientation, and location of the pattern. This independence can be achieved by preprocessing or by extracting features that are translation-, rotation-, and scale-invariant.