Shape Analysis and Classification using Landmarks: Polygonal Wavelet Transform

Shape analysis has played a central role in many problems in vision and perception, being an active multidisciplinary research field. In this context, this paper introduces a new shape analysis approach using the well known wavelet transform and exploring shape representation by landmarks. This work shows how to obtain a time signal from the landmarks representation, which undergoes the wavelet transform, as well as a useful geometrical interpretation using two special mother wavelets, i.e. the first and the second derivatives of the gaussian. Successful experimental results obtained from real data are also discussed.