Feature extraction using wavelet and fractal

In this paper, we are investigating the utility of several emerging techniq ues to extract features. A novel method of feature extraction is proposed, which includes utilizing the central projection transformation (CPT) to des cribe the shape, the wavelet transformation to aid in the boundary identifi cation, and the fractal features to enhance image discrimination. It reduce s the dimensionality of a two-dimensional pattern by way of a central proje ction approach, and thereafter, performs Daubechies' wavelet transform on t he derived one-dimensional pattern to generate a set of wavelet transform s ub-patterns, namely, curves that are non-self-intersecting. The divider dim ensions are computed from these curves with a modified box-counting approac h. These divider dimensions constitute a new feature vector for the origina l two-dimensional pattern, defined over the curve's fractal dimensions. We have conducted several experiments in which a set of printed Chinese charac ters, English letters of varying fonts and other images were classified. Ba sed on the Euclidean distance between the different feature vectors, the ex periments have satisfying results. (C) 2001 Elsevier Science B.V. All right s reserved.